diff --git a/docs/configuration.md b/docs/configuration.md
index 3462767e6..9c201c787 100644
--- a/docs/configuration.md
+++ b/docs/configuration.md
@@ -246,6 +246,88 @@ In other words: **channel + account form the candidate set; peer/guild/team then
 - **Wildcard catches too much traffic?** Add more specific `peer/guild/team` rules for critical paths.
 - **Unexpected default fallback?** Confirm `agent_id` exists and is not misspelled.
 
+### Agent Tool Allowlist
+
+You can restrict an individual agent to a subset of runtime tools with `agents.list[].tools`.
+
+If `tools` is omitted, the agent gets the normal globally enabled tool set. If `tools` is present, PicoClaw registers only the listed tools for that agent.
+
+```json
+{
+  "agents": {
+    "list": [
+      {
+        "id": "research",
+        "name": "Research Agent",
+        "tools": ["read_file", "write_file", "web_search", "web_fetch", "message"]
+      }
+    ]
+  }
+}
+```
+
+Notes:
+
+- This is an allowlist, not a preference hint.
+- Tool names are matched against the runtime tool name 1:1.
+- Use runtime tool names such as `web_search`, `web_fetch`, `spawn`, `subagent`, `send_file`.
+- The `available_tools` field in Agent Discovery reflects the filtered runtime result.
+
+### Agent Discovery (Automatic)
+
+When more than one agent exists, PicoClaw injects a structured agent registry into each agent's system prompt on every turn. No extra `list_agents` tool call is required.
+
+This registry is intended to make delegation concrete and reliable, especially when using `spawn` with a target `agent_id`.
+
+Each entry includes:
+
+| Field | Meaning |
+|-------|---------|
+| `id` | Stable agent id |
+| `name` | Human-friendly agent name |
+| `description` | Short capability summary |
+| `model` | Current model used by that agent |
+| `available_tools` | Tool names currently visible to that agent |
+| `channels` | Channels that route to that agent |
+
+Important behavior:
+
+- The discovery section includes the current agent's own entry, so the model has self-awareness.
+- `available_tools` is the most important field for delegation. It reflects the tools the target agent can actually use, not just a natural-language description.
+- `description` is sourced from `AGENT.md` frontmatter `description` when available, otherwise from the first meaningful paragraph of `AGENT.md`, and finally `SOUL.md`.
+- `name` comes from `agents.list[].name` first, then `AGENT.md` frontmatter `name`, then falls back to the agent id.
+- `channels` come from routing state:
+  - the default agent exposes enabled channels
+  - other agents expose channels that explicitly bind to them through `bindings`
+
+Example injected shape:
+
+```json
+{
+  "current_agent_id": "main",
+  "agents": [
+    {
+      "id": "main",
+      "name": "Main Assistant",
+      "description": "Generalist agent for day-to-day requests.",
+      "model": "gpt-4o-mini",
+      "available_tools": ["read_file", "write_file", "exec", "spawn"],
+      "channels": ["telegram", "discord"]
+    },
+    {
+      "id": "research",
+      "name": "Research Agent",
+      "description": "Specialist for long-form investigation and web work.",
+      "model": "claude-sonnet-4.5",
+      "available_tools": ["web_search", "web_fetch", "read_file"],
+      "channels": ["telegram"]
+    }
+  ]
+}
+```
+
+In practice, this means a generalist agent can see that a peer has `["web_search", "web_fetch"]` while it only has local file tools, and can decide to delegate to that peer instead of guessing.
+
 ### 🔒 Security Sandbox
 
 PicoClaw runs in a sandboxed environment by default. The agent can only access files and execute commands within the configured workspace.
diff --git a/docs/it/configuration.md b/docs/it/configuration.md
index 6a79a9543..9b0d4a198 100644
--- a/docs/it/configuration.md
+++ b/docs/it/configuration.md
@@ -42,14 +42,13 @@ PicoClaw salva i dati nel workspace configurato (predefinito: `~/.picoclaw/works
 ├── state/            # Stato persistente (ultimo canale, ecc.)
 ├── cron/             # Database dei job pianificati
 ├── skills/           # Skill personalizzate
-├── AGENTS.md         # Guida al comportamento dell'agent
+├── AGENT.md          # Guida al comportamento dell'agent
 ├── HEARTBEAT.md      # Prompt per task periodici (controllato ogni 30 min)
-├── IDENTITY.md       # Identità dell'agent
 ├── SOUL.md           # Anima dell'agent
 └── USER.md           # Preferenze dell'utente
 ```
 
-> **Nota:** Le modifiche a `AGENTS.md`, `SOUL.md`, `USER.md`, `IDENTITY.md` e `memory/MEMORY.md` vengono rilevate automaticamente a runtime tramite il tracciamento della data di modifica (mtime). **Non è necessario riavviare il gateway** dopo aver modificato questi file — l'agent caricherà il nuovo contenuto alla prossima richiesta.
+> **Nota:** Le modifiche a `AGENT.md`, `SOUL.md`, `USER.md` e `memory/MEMORY.md` vengono rilevate automaticamente a runtime tramite il tracciamento della data di modifica (mtime). **Non è necessario riavviare il gateway** dopo aver modificato questi file — l'agent caricherà il nuovo contenuto alla prossima richiesta.
 
 ### Sorgenti delle Skill
 
@@ -72,6 +71,88 @@ export PICOCLAW_BUILTIN_SKILLS=/path/to/skills
 - Un comando slash sconosciuto (ad esempio `/foo`) viene passato all'elaborazione LLM come se fosse un messaggio dell'utente.
 - Un comando registrato ma non supportato sul canale corrente (ad esempio `/show` su WhatsApp) restituisce un errore esplicito all'utente e interrompe l'elaborazione.
 
+### Allowlist dei Tool per Agent
+
+Puoi limitare un singolo agent a un sottoinsieme di tool runtime con `agents.list[].tools`.
+
+Se `tools` è omesso, l'agent riceve il normale set globale dei tool abilitati. Se `tools` è presente, PicoClaw registra per quell'agent solo i tool elencati.
+
+```json
+{
+  "agents": {
+    "list": [
+      {
+        "id": "research",
+        "name": "Research Agent",
+        "tools": ["read_file", "write_file", "web_search", "web_fetch", "message"]
+      }
+    ]
+  }
+}
+```
+
+Note:
+
+- È una allowlist reale, non un suggerimento per l'LLM.
+- I nomi dei tool fanno match 1:1 con il nome runtime del tool.
+- Se ti serve controllo preciso, usa i nomi runtime effettivi come `web_search`, `web_fetch`, `spawn`, `subagent`, `send_file`.
+- Il campo `available_tools` nella Agent Discovery riflette il risultato filtrato reale.
+
+### Discovery Multi-Agent (Automatica)
+
+Quando esiste più di un agent, PicoClaw inietta automaticamente nel system prompt di ogni agent un registry strutturato dei peer. Non serve una chiamata aggiuntiva a un tool `list_agents`.
+
+Questa discovery serve soprattutto a rendere affidabile la delega tramite `spawn` con `agent_id` esplicito.
+
+Ogni entry include:
+
+| Campo | Significato |
+|-------|-------------|
+| `id` | ID stabile dell'agent |
+| `name` | Nome leggibile dell'agent |
+| `description` | Riassunto breve delle capacità |
+| `model` | Modello attualmente usato da quell'agent |
+| `available_tools` | Tool attualmente visibili a quell'agent |
+| `channels` | Canali instradati verso quell'agent |
+
+Dettagli importanti:
+
+- La sezione include anche l'entry dell'agent corrente, quindi c'è self-awareness.
+- `available_tools` è il campo più importante per delegare bene: l'LLM vede i tool reali del peer, non deve indovinarli dalla sola descrizione.
+- `description` viene presa da `AGENT.md` frontmatter `description` quando presente; altrimenti dal primo paragrafo utile di `AGENT.md`, e in fallback da `SOUL.md`.
+- `name` arriva prima da `agents.list[].name`, poi da `AGENT.md` frontmatter `name`, e in fallback dall'ID dell'agent.
+- `channels` derivano dal routing:
+  - l'agent di default espone i canali abilitati
+  - gli altri agent espongono i canali che hanno un binding esplicito verso di loro
+
+Forma dell'oggetto iniettato:
+
+```json
+{
+  "current_agent_id": "main",
+  "agents": [
+    {
+      "id": "main",
+      "name": "Main Assistant",
+      "description": "Agent generalista per richieste quotidiane.",
+      "model": "gpt-4o-mini",
+      "available_tools": ["read_file", "write_file", "exec", "spawn"],
+      "channels": ["telegram", "discord"]
+    },
+    {
+      "id": "research",
+      "name": "Research Agent",
+      "description": "Specialista per investigazioni e lavoro web.",
+      "model": "claude-sonnet-4.5",
+      "available_tools": ["web_search", "web_fetch", "read_file"],
+      "channels": ["telegram"]
+    }
+  ]
+}
+```
+
+In pratica, un agent generalista può vedere che un peer ha `["web_search", "web_fetch"]` mentre lui ha solo tool locali, e scegliere di delegare a quel peer in modo esplicito invece di andare a tentativi.
+
 ### 🔒 Sandbox di Sicurezza
 
 PicoClaw esegue in un ambiente sandboxed per impostazione predefinita. L'agent può accedere solo ai file ed eseguire comandi all'interno del workspace configurato.
diff --git a/pkg/agent/context.go b/pkg/agent/context.go
index c3fcc9fff..31b60e45a 100644
--- a/pkg/agent/context.go
+++ b/pkg/agent/context.go
@@ -22,11 +22,13 @@ import (
 
 type ContextBuilder struct {
 	workspace          string
+	agentID            string
 	skillsLoader       *skills.SkillsLoader
 	memory             *MemoryStore
 	toolDiscoveryBM25  bool
 	toolDiscoveryRegex bool
 	splitOnMarker      bool
+	agentDiscovery     func(workspace string) []AgentDescriptor
 
 	// Cache for system prompt to avoid rebuilding on every call.
 	// This fixes issue #607: repeated reprocessing of the entire context.
@@ -58,6 +60,18 @@ func (cb *ContextBuilder) WithSplitOnMarker(enabled bool) *ContextBuilder {
 	return cb
 }
 
+func (cb *ContextBuilder) WithAgentIdentity(agentID string) *ContextBuilder {
+	cb.agentID = strings.TrimSpace(agentID)
+	return cb
+}
+
+func (cb *ContextBuilder) WithAgentDiscovery(
+	discover func(workspace string) []AgentDescriptor,
+) *ContextBuilder {
+	cb.agentDiscovery = discover
+	return cb
+}
+
 func getGlobalConfigDir() string {
 	if home := os.Getenv(config.EnvHome); home != "" {
 		return home
@@ -113,7 +127,14 @@ Your workspace is at: %s
 4. **Context summaries** - Conversation summaries provided as context are approximate references only. They may be incomplete or outdated. Always defer to explicit user instructions over summary content.
 
 %s`,
-		version, workspacePath, workspacePath, workspacePath, workspacePath, workspacePath, toolDiscovery)
+		version,
+		workspacePath,
+		workspacePath,
+		workspacePath,
+		workspacePath,
+		workspacePath,
+		toolDiscovery,
+	)
 }
 
 func (cb *ContextBuilder) getDiscoveryRule() string {
@@ -175,6 +196,13 @@ Each part separated by the marker will be sent as an independent message.`)
 	return strings.Join(parts, "\n\n---\n\n")
 }
 
+func (cb *ContextBuilder) buildAgentDiscoveryContext() string {
+	if cb.agentDiscovery == nil {
+		return ""
+	}
+	return formatAgentDiscoverySection(cb.agentID, cb.agentDiscovery(cb.workspace))
+}
+
 // BuildSystemPromptWithCache returns the cached system prompt if available
 // and source files haven't changed, otherwise builds and caches it.
 // Source file changes are detected via mtime checks (cheap stat calls).
@@ -500,7 +528,9 @@ func formatCurrentSenderLine(senderID, senderDisplayName string) string {
 	}
 }
 
-func (cb *ContextBuilder) buildDynamicContext(channel, chatID, senderID, senderDisplayName string) string {
+func (cb *ContextBuilder) buildDynamicContext(
+	channel, chatID, senderID, senderDisplayName string,
+) string {
 	now := time.Now().Format("2006-01-02 15:04 (Monday)")
 	rt := fmt.Sprintf("%s %s, Go %s", runtime.GOOS, runtime.GOARCH, runtime.Version())
 
@@ -540,6 +570,7 @@ func (cb *ContextBuilder) BuildMessages(
 
 	// Build short dynamic context (time, runtime, session) — changes per request
 	dynamicCtx := cb.buildDynamicContext(channel, chatID, senderID, senderDisplayName)
+	discoveryCtx := cb.buildAgentDiscoveryContext()
 
 	// Compose a single system message: static (cached) + dynamic + optional summary.
 	// Keeping all system content in one message ensures every provider adapter can
@@ -550,16 +581,33 @@ func (cb *ContextBuilder) BuildMessages(
 	// cache-aware adapters (Anthropic) can set per-block cache_control.
 	// The static block is marked "ephemeral" — its prefix hash is stable
 	// across requests, enabling LLM-side KV cache reuse.
-	stringParts := []string{staticPrompt, dynamicCtx}
+	stringParts := []string{staticPrompt}
 
 	contentBlocks := []providers.ContentBlock{
-		{Type: "text", Text: staticPrompt, CacheControl: &providers.CacheControl{Type: "ephemeral"}},
-		{Type: "text", Text: dynamicCtx},
+		{
+			Type:         "text",
+			Text:         staticPrompt,
+			CacheControl: &providers.CacheControl{Type: "ephemeral"},
+		},
 	}
 
+	if discoveryCtx != "" {
+		stringParts = append(stringParts, discoveryCtx)
+		contentBlocks = append(
+			contentBlocks,
+			providers.ContentBlock{Type: "text", Text: discoveryCtx},
+		)
+	}
+
+	stringParts = append(stringParts, dynamicCtx)
+	contentBlocks = append(contentBlocks, providers.ContentBlock{Type: "text", Text: dynamicCtx})
+
 	if skillsText := cb.buildActiveSkillsContext(activeSkills); skillsText != "" {
 		stringParts = append(stringParts, skillsText)
-		contentBlocks = append(contentBlocks, providers.ContentBlock{Type: "text", Text: skillsText})
+		contentBlocks = append(
+			contentBlocks,
+			providers.ContentBlock{Type: "text", Text: skillsText},
+		)
 	}
 
 	if summary != "" {
@@ -568,7 +616,10 @@ func (cb *ContextBuilder) BuildMessages(
 				"for reference only. It may be incomplete or outdated — always defer to explicit instructions.\n\n%s",
 			summary)
 		stringParts = append(stringParts, summaryText)
-		contentBlocks = append(contentBlocks, providers.ContentBlock{Type: "text", Text: summaryText})
+		contentBlocks = append(
+			contentBlocks,
+			providers.ContentBlock{Type: "text", Text: summaryText},
+		)
 	}
 
 	fullSystemPrompt := strings.Join(stringParts, "\n\n---\n\n")
@@ -667,7 +718,11 @@ func sanitizeHistoryForProvider(history []providers.Message) []providers.Message
 		case "assistant":
 			if len(msg.ToolCalls) > 0 {
 				if len(sanitized) == 0 {
-					logger.DebugCF("agent", "Dropping assistant tool-call turn at history start", map[string]any{})
+					logger.DebugCF(
+						"agent",
+						"Dropping assistant tool-call turn at history start",
+						map[string]any{},
+					)
 					continue
 				}
 				prev := sanitized[len(sanitized)-1]
diff --git a/pkg/agent/context_budget_test.go b/pkg/agent/context_budget_test.go
index 870f0fbe6..c8993746f 100644
--- a/pkg/agent/context_budget_test.go
+++ b/pkg/agent/context_budget_test.go
@@ -500,8 +500,11 @@ func TestEstimateMessageTokens_ReasoningContent(t *testing.T) {
 	reasoningTokens := estimateMessageTokens(withReasoning)
 
 	if reasoningTokens <= plainTokens {
-		t.Errorf("message with ReasoningContent (%d tokens) should exceed plain message (%d tokens)",
-			reasoningTokens, plainTokens)
+		t.Errorf(
+			"message with ReasoningContent (%d tokens) should exceed plain message (%d tokens)",
+			reasoningTokens,
+			plainTokens,
+		)
 	}
 }
 
@@ -764,7 +767,11 @@ func TestEstimateMessageTokens_WithReasoningAndMedia(t *testing.T) {
 	tokensNoReasoning := estimateMessageTokens(msgNoReasoning)
 
 	if tokens <= tokensNoReasoning {
-		t.Errorf("reasoning content should add tokens: with=%d, without=%d", tokens, tokensNoReasoning)
+		t.Errorf(
+			"reasoning content should add tokens: with=%d, without=%d",
+			tokens,
+			tokensNoReasoning,
+		)
 	}
 }
 
diff --git a/pkg/agent/context_cache_test.go b/pkg/agent/context_cache_test.go
index 81a1534b9..ae6ff18cc 100644
--- a/pkg/agent/context_cache_test.go
+++ b/pkg/agent/context_cache_test.go
@@ -82,7 +82,16 @@ func TestSingleSystemMessage(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			msgs := cb.BuildMessages(tt.history, tt.summary, tt.message, nil, "test", "chat1", "", "")
+			msgs := cb.BuildMessages(
+				tt.history,
+				tt.summary,
+				tt.message,
+				nil,
+				"test",
+				"chat1",
+				"",
+				"",
+			)
 
 			systemCount := 0
 			for _, m := range msgs {
@@ -168,7 +177,16 @@ func TestBuildMessages_CurrentSenderDynamicContext(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			msgs := cb.BuildMessages(nil, "", "hello", nil, "discord", "chat1", tt.senderID, tt.senderDisplayName)
+			msgs := cb.BuildMessages(
+				nil,
+				"",
+				"hello",
+				nil,
+				"discord",
+				"chat1",
+				tt.senderID,
+				tt.senderDisplayName,
+			)
 			sys := msgs[0].Content
 
 			if tt.wantSection {
@@ -382,7 +400,10 @@ func TestNewFileCreationInvalidatesCache(t *testing.T) {
 			// Cache should auto-invalidate because file went from absent -> present
 			sp2 := cb.BuildSystemPromptWithCache()
 			if !strings.Contains(sp2, tt.checkField) {
-				t.Errorf("cache not invalidated on new file creation: expected %q in prompt", tt.checkField)
+				t.Errorf(
+					"cache not invalidated on new file creation: expected %q in prompt",
+					tt.checkField,
+				)
 			}
 		})
 	}
diff --git a/pkg/agent/context_test.go b/pkg/agent/context_test.go
index 0d7948eef..c3b9ed6a0 100644
--- a/pkg/agent/context_test.go
+++ b/pkg/agent/context_test.go
@@ -151,7 +151,19 @@ func TestSanitizeHistoryForProvider_MultiToolCallsThenNewRound(t *testing.T) {
 	if len(result) != 9 {
 		t.Fatalf("expected 9 messages, got %d: %+v", len(result), roles(result))
 	}
-	assertRoles(t, result, "user", "assistant", "tool", "tool", "assistant", "user", "assistant", "tool", "assistant")
+	assertRoles(
+		t,
+		result,
+		"user",
+		"assistant",
+		"tool",
+		"tool",
+		"assistant",
+		"user",
+		"assistant",
+		"tool",
+		"assistant",
+	)
 }
 
 func TestSanitizeHistoryForProvider_ConsecutiveMultiToolRounds(t *testing.T) {
@@ -170,7 +182,18 @@ func TestSanitizeHistoryForProvider_ConsecutiveMultiToolRounds(t *testing.T) {
 	if len(result) != 8 {
 		t.Fatalf("expected 8 messages, got %d: %+v", len(result), roles(result))
 	}
-	assertRoles(t, result, "user", "assistant", "tool", "tool", "assistant", "tool", "tool", "assistant")
+	assertRoles(
+		t,
+		result,
+		"user",
+		"assistant",
+		"tool",
+		"tool",
+		"assistant",
+		"tool",
+		"tool",
+		"assistant",
+	)
 }
 
 func TestSanitizeHistoryForProvider_PlainConversation(t *testing.T) {
@@ -304,5 +327,17 @@ func TestSanitizeHistoryForProvider_PartialToolResultsInMiddle(t *testing.T) {
 	if len(result) != 9 {
 		t.Fatalf("expected 9 messages, got %d: %+v", len(result), roles(result))
 	}
-	assertRoles(t, result, "user", "assistant", "tool", "assistant", "user", "user", "assistant", "tool", "assistant")
+	assertRoles(
+		t,
+		result,
+		"user",
+		"assistant",
+		"tool",
+		"assistant",
+		"user",
+		"user",
+		"assistant",
+		"tool",
+		"assistant",
+	)
 }
diff --git a/pkg/agent/definition_test.go b/pkg/agent/definition_test.go
index 5ee996967..b3068d134 100644
--- a/pkg/agent/definition_test.go
+++ b/pkg/agent/definition_test.go
@@ -61,8 +61,12 @@ Act directly and use tools first.
 	if len(definition.Agent.Frontmatter.Skills) != 2 {
 		t.Fatalf("expected skills to be parsed, got %v", definition.Agent.Frontmatter.Skills)
 	}
-	if len(definition.Agent.Frontmatter.MCPServers) != 1 || definition.Agent.Frontmatter.MCPServers[0] != "github" {
-		t.Fatalf("expected mcpServers to be parsed, got %v", definition.Agent.Frontmatter.MCPServers)
+	if len(definition.Agent.Frontmatter.MCPServers) != 1 ||
+		definition.Agent.Frontmatter.MCPServers[0] != "github" {
+		t.Fatalf(
+			"expected mcpServers to be parsed, got %v",
+			definition.Agent.Frontmatter.MCPServers,
+		)
 	}
 	if definition.Agent.Frontmatter.Fields["metadata"] == nil {
 		t.Fatal("expected arbitrary frontmatter fields to remain available")
@@ -96,7 +100,10 @@ func TestLoadAgentDefinitionFallsBackToLegacyAgentsMarkdown(t *testing.T) {
 		t.Fatal("expected AGENTS.md to be loaded")
 	}
 	if definition.Agent.RawFrontmatter != "" {
-		t.Fatalf("legacy AGENTS.md should not have frontmatter, got %q", definition.Agent.RawFrontmatter)
+		t.Fatalf(
+			"legacy AGENTS.md should not have frontmatter, got %q",
+			definition.Agent.RawFrontmatter,
+		)
 	}
 	if !strings.Contains(definition.Agent.Body, "Keep compatibility") {
 		t.Fatalf("expected legacy body to be preserved, got %q", definition.Agent.Body)
@@ -159,7 +166,10 @@ Keep going.
 		len(definition.Agent.Frontmatter.Skills) != 0 ||
 		len(definition.Agent.Frontmatter.MCPServers) != 0 ||
 		len(definition.Agent.Frontmatter.Fields) != 0 {
-		t.Fatalf("expected invalid frontmatter to decode as empty struct, got %+v", definition.Agent.Frontmatter)
+		t.Fatalf(
+			"expected invalid frontmatter to decode as empty struct, got %+v",
+			definition.Agent.Frontmatter,
+		)
 	}
 }
 
diff --git a/pkg/agent/discovery.go b/pkg/agent/discovery.go
new file mode 100644
index 000000000..b630abd60
--- /dev/null
+++ b/pkg/agent/discovery.go
@@ -0,0 +1,341 @@
+package agent
+
+import (
+	"encoding/json"
+	"fmt"
+	"path/filepath"
+	"sort"
+	"strings"
+
+	"github.com/sipeed/picoclaw/pkg/config"
+	"github.com/sipeed/picoclaw/pkg/routing"
+)
+
+// AgentDescriptor is the structured discovery payload injected into each
+// agent's system prompt so the LLM can make concrete delegation decisions.
+type AgentDescriptor struct {
+	ID             string   `json:"id"`
+	Name           string   `json:"name"`
+	Description    string   `json:"description"`
+	Model          string   `json:"model"`
+	AvailableTools []string `json:"available_tools"`
+	Channels       []string `json:"channels"`
+}
+
+// ListAgents returns structured descriptors for every agent in the current
+// PicoClaw instance. The current workspace, when provided, is used only to
+// order the matching agent first for prompt readability.
+func (r *AgentRegistry) ListAgents(workspace string) []AgentDescriptor {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	ids := make([]string, 0, len(r.agents))
+	for id := range r.agents {
+		ids = append(ids, id)
+	}
+	sort.Strings(ids)
+
+	selfWorkspace := cleanWorkspacePath(workspace)
+	descriptors := make([]AgentDescriptor, 0, len(ids))
+	for _, id := range ids {
+		agent := r.agents[id]
+		if agent == nil {
+			continue
+		}
+		descriptors = append(descriptors, r.buildAgentDescriptorLocked(agent))
+	}
+
+	if selfWorkspace == "" {
+		return descriptors
+	}
+
+	sort.SliceStable(descriptors, func(i, j int) bool {
+		leftSelf := cleanWorkspacePath(
+			r.workspaceForAgentIDLocked(descriptors[i].ID),
+		) == selfWorkspace
+		rightSelf := cleanWorkspacePath(
+			r.workspaceForAgentIDLocked(descriptors[j].ID),
+		) == selfWorkspace
+		if leftSelf != rightSelf {
+			return leftSelf
+		}
+		return descriptors[i].ID < descriptors[j].ID
+	})
+
+	return descriptors
+}
+
+// GetAgentDescriptor returns the structured discovery payload for one agent.
+func (r *AgentRegistry) GetAgentDescriptor(agentID string) (*AgentDescriptor, bool) {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	id := routing.NormalizeAgentID(agentID)
+	agent, ok := r.agents[id]
+	if !ok || agent == nil {
+		return nil, false
+	}
+
+	descriptor := r.buildAgentDescriptorLocked(agent)
+	return &descriptor, true
+}
+
+func (r *AgentRegistry) buildAgentDescriptorLocked(agent *AgentInstance) AgentDescriptor {
+	definition := loadAgentDefinition(agent.Workspace)
+	name := strings.TrimSpace(agent.Name)
+	if name == "" && definition.Agent != nil {
+		name = strings.TrimSpace(definition.Agent.Frontmatter.Name)
+	}
+	if name == "" {
+		name = agent.ID
+	}
+
+	return AgentDescriptor{
+		ID:             agent.ID,
+		Name:           name,
+		Description:    agentDescriptionFromDefinition(definition),
+		Model:          strings.TrimSpace(agent.Model),
+		AvailableTools: visibleToolNames(agent),
+		Channels:       r.channelsForAgentLocked(agent.ID),
+	}
+}
+
+func visibleToolNames(agent *AgentInstance) []string {
+	if agent == nil || agent.Tools == nil {
+		return []string{}
+	}
+
+	defs := agent.Tools.ToProviderDefs()
+	names := make([]string, 0, len(defs))
+	for _, def := range defs {
+		name := strings.TrimSpace(def.Function.Name)
+		if name == "" {
+			continue
+		}
+		names = append(names, name)
+	}
+	if names == nil {
+		return []string{}
+	}
+	return names
+}
+
+func agentDescriptionFromDefinition(definition AgentContextDefinition) string {
+	if definition.Agent != nil {
+		if desc := strings.TrimSpace(definition.Agent.Frontmatter.Description); desc != "" {
+			return desc
+		}
+		if desc := firstMeaningfulParagraph(definition.Agent.Body); desc != "" {
+			return desc
+		}
+	}
+	if definition.Soul != nil {
+		if desc := firstMeaningfulParagraph(definition.Soul.Content); desc != "" {
+			return desc
+		}
+	}
+	return ""
+}
+
+func firstMeaningfulParagraph(content string) string {
+	content = strings.ReplaceAll(content, "\r\n", "\n")
+	paragraphs := strings.Split(content, "\n\n")
+	for _, paragraph := range paragraphs {
+		lines := strings.Split(paragraph, "\n")
+		parts := make([]string, 0, len(lines))
+		inFence := false
+		for _, line := range lines {
+			trimmed := strings.TrimSpace(line)
+			if strings.HasPrefix(trimmed, "```") {
+				inFence = !inFence
+				continue
+			}
+			if inFence || trimmed == "" {
+				continue
+			}
+			if strings.HasPrefix(trimmed, "#") {
+				continue
+			}
+			if strings.HasPrefix(trimmed, "- ") || strings.HasPrefix(trimmed, "* ") {
+				trimmed = strings.TrimSpace(trimmed[2:])
+			}
+			parts = append(parts, trimmed)
+		}
+		if len(parts) == 0 {
+			continue
+		}
+		return strings.Join(parts, " ")
+	}
+	return ""
+}
+
+func (r *AgentRegistry) channelsForAgentLocked(agentID string) []string {
+	channels := make(map[string]struct{})
+
+	if defaultID := r.defaultAgentIDLocked(); defaultID != "" && defaultID == agentID {
+		for _, channel := range enabledChannels(r.cfg) {
+			channels[channel] = struct{}{}
+		}
+	}
+
+	if r.cfg != nil {
+		for _, binding := range r.cfg.Bindings {
+			if routing.NormalizeAgentID(binding.AgentID) != agentID {
+				continue
+			}
+			channel := strings.ToLower(strings.TrimSpace(binding.Match.Channel))
+			if channel == "" {
+				continue
+			}
+			channels[channel] = struct{}{}
+		}
+	}
+
+	if len(channels) == 0 {
+		return []string{}
+	}
+
+	result := make([]string, 0, len(channels))
+	for channel := range channels {
+		result = append(result, channel)
+	}
+	sort.Strings(result)
+	return result
+}
+
+func enabledChannels(cfg *config.Config) []string {
+	if cfg == nil {
+		return []string{}
+	}
+
+	enabled := make([]string, 0, 16)
+	if cfg.Channels.WhatsApp.Enabled {
+		enabled = append(enabled, "whatsapp")
+	}
+	if cfg.Channels.Telegram.Enabled {
+		enabled = append(enabled, "telegram")
+	}
+	if cfg.Channels.Feishu.Enabled {
+		enabled = append(enabled, "feishu")
+	}
+	if cfg.Channels.Discord.Enabled {
+		enabled = append(enabled, "discord")
+	}
+	if cfg.Channels.MaixCam.Enabled {
+		enabled = append(enabled, "maixcam")
+	}
+	if cfg.Channels.QQ.Enabled {
+		enabled = append(enabled, "qq")
+	}
+	if cfg.Channels.DingTalk.Enabled {
+		enabled = append(enabled, "dingtalk")
+	}
+	if cfg.Channels.Slack.Enabled {
+		enabled = append(enabled, "slack")
+	}
+	if cfg.Channels.Matrix.Enabled {
+		enabled = append(enabled, "matrix")
+	}
+	if cfg.Channels.LINE.Enabled {
+		enabled = append(enabled, "line")
+	}
+	if cfg.Channels.OneBot.Enabled {
+		enabled = append(enabled, "onebot")
+	}
+	if cfg.Channels.WeCom.Enabled {
+		enabled = append(enabled, "wecom")
+	}
+	if cfg.Channels.Weixin.Enabled {
+		enabled = append(enabled, "weixin")
+	}
+	if cfg.Channels.Pico.Enabled {
+		enabled = append(enabled, "pico")
+	}
+	if cfg.Channels.PicoClient.Enabled {
+		enabled = append(enabled, "pico_client")
+	}
+	if cfg.Channels.IRC.Enabled {
+		enabled = append(enabled, "irc")
+	}
+	return enabled
+}
+
+func (r *AgentRegistry) workspaceForAgentIDLocked(agentID string) string {
+	agent, ok := r.agents[routing.NormalizeAgentID(agentID)]
+	if !ok || agent == nil {
+		return ""
+	}
+	return agent.Workspace
+}
+
+func (r *AgentRegistry) defaultAgentIDLocked() string {
+	if _, ok := r.agents[routing.DefaultAgentID]; ok {
+		return routing.DefaultAgentID
+	}
+	if r.cfg != nil && len(r.cfg.Agents.List) > 0 {
+		for _, agentCfg := range r.cfg.Agents.List {
+			if !agentCfg.Default {
+				continue
+			}
+			id := routing.NormalizeAgentID(agentCfg.ID)
+			if _, ok := r.agents[id]; ok {
+				return id
+			}
+		}
+		id := routing.NormalizeAgentID(r.cfg.Agents.List[0].ID)
+		if _, ok := r.agents[id]; ok {
+			return id
+		}
+	}
+	for id := range r.agents {
+		return id
+	}
+	return ""
+}
+
+func cleanWorkspacePath(path string) string {
+	path = strings.TrimSpace(path)
+	if path == "" {
+		return ""
+	}
+	return filepath.Clean(path)
+}
+
+func formatAgentDiscoverySection(currentAgentID string, agents []AgentDescriptor) string {
+	if len(agents) <= 1 {
+		return ""
+	}
+
+	payload := struct {
+		CurrentAgentID string            `json:"current_agent_id"`
+		Agents         []AgentDescriptor `json:"agents"`
+	}{
+		CurrentAgentID: strings.TrimSpace(currentAgentID),
+		Agents:         agents,
+	}
+
+	encoded, err := json.MarshalIndent(payload, "", "  ")
+	if err != nil {
+		return ""
+	}
+
+	var header strings.Builder
+	header.WriteString("# Agent Discovery\n\n")
+	if payload.CurrentAgentID != "" {
+		fmt.Fprintf(
+			&header,
+			"You are agent %q. This registry is authoritative for the current PicoClaw instance and includes your own entry.\n",
+			payload.CurrentAgentID,
+		)
+	} else {
+		header.WriteString("This registry is authoritative for the current PicoClaw instance.\n")
+	}
+	header.WriteString(
+		"Delegate based on available_tools first, then model, channels, and description. Use only agent IDs listed here.\n\n",
+	)
+	header.WriteString("```json\n")
+	header.Write(encoded)
+	header.WriteString("\n```")
+
+	return header.String()
+}
diff --git a/pkg/agent/discovery_test.go b/pkg/agent/discovery_test.go
new file mode 100644
index 000000000..a44f67dea
--- /dev/null
+++ b/pkg/agent/discovery_test.go
@@ -0,0 +1,211 @@
+package agent
+
+import (
+	"slices"
+	"strings"
+	"testing"
+
+	"github.com/sipeed/picoclaw/pkg/config"
+)
+
+func TestAgentRegistry_ListAgentsBuildsStructuredDescriptors(t *testing.T) {
+	mainWorkspace := setupWorkspace(t, map[string]string{
+		"AGENT.md": `---
+name: Main Frontmatter Name
+description: Structured main agent
+---
+# Agent
+
+Handle general requests.
+`,
+	})
+	defer cleanupWorkspace(t, mainWorkspace)
+
+	supportWorkspace := setupWorkspace(t, map[string]string{
+		"AGENT.md": `# Agent
+
+Handle support tickets carefully.
+`,
+		"SOUL.md": "# Soul\nStay calm and precise.",
+	})
+	defer cleanupWorkspace(t, supportWorkspace)
+
+	cfg := testCfg([]config.AgentConfig{
+		{ID: "main", Default: true, Name: "Configured Main", Workspace: mainWorkspace},
+		{
+			ID:        "support",
+			Workspace: supportWorkspace,
+			Model:     &config.AgentModelConfig{Primary: "support-model"},
+		},
+	})
+	cfg.Tools.ReadFile.Enabled = true
+	cfg.Tools.WriteFile.Enabled = true
+	cfg.Channels.Telegram.Enabled = true
+	cfg.Bindings = []config.AgentBinding{
+		{
+			AgentID: "support",
+			Match: config.BindingMatch{
+				Channel:   "telegram",
+				AccountID: "*",
+			},
+		},
+	}
+
+	registry := NewAgentRegistry(cfg, &mockRegistryProvider{})
+
+	descriptors := registry.ListAgents(mainWorkspace)
+	if len(descriptors) != 2 {
+		t.Fatalf("expected 2 descriptors, got %d", len(descriptors))
+	}
+
+	if descriptors[0].ID != "main" {
+		t.Fatalf("expected current workspace agent first, got %q", descriptors[0].ID)
+	}
+	if descriptors[0].Name != "Configured Main" {
+		t.Fatalf("expected config name to win, got %q", descriptors[0].Name)
+	}
+	if descriptors[0].Description != "Structured main agent" {
+		t.Fatalf("expected frontmatter description, got %q", descriptors[0].Description)
+	}
+	if descriptors[0].Model != "gpt-4" {
+		t.Fatalf("expected inherited model, got %q", descriptors[0].Model)
+	}
+	if !slices.Contains(descriptors[0].AvailableTools, "read_file") ||
+		!slices.Contains(descriptors[0].AvailableTools, "write_file") {
+		t.Fatalf("expected visible file tools in descriptor, got %v", descriptors[0].AvailableTools)
+	}
+	if !slices.Equal(descriptors[0].Channels, []string{"telegram"}) {
+		t.Fatalf(
+			"expected default agent to cover enabled telegram channel, got %v",
+			descriptors[0].Channels,
+		)
+	}
+
+	support, ok := registry.GetAgentDescriptor("support")
+	if !ok || support == nil {
+		t.Fatal("expected support descriptor lookup to succeed")
+	}
+	if support.Description != "Handle support tickets carefully." {
+		t.Fatalf("expected AGENT body fallback description, got %q", support.Description)
+	}
+	if support.Model != "support-model" {
+		t.Fatalf("expected explicit support model, got %q", support.Model)
+	}
+	if !slices.Equal(support.Channels, []string{"telegram"}) {
+		t.Fatalf("expected support channel binding, got %v", support.Channels)
+	}
+}
+
+func TestContextBuilder_BuildMessagesIncludesAgentDiscoverySection(t *testing.T) {
+	mainWorkspace := setupWorkspace(t, map[string]string{
+		"AGENT.md": `---
+description: Main agent
+---
+# Agent
+
+Generalist.
+`,
+	})
+	defer cleanupWorkspace(t, mainWorkspace)
+
+	researchWorkspace := setupWorkspace(t, map[string]string{
+		"AGENT.md": `---
+description: Research specialist
+---
+# Agent
+
+Investigate deeply.
+`,
+	})
+	defer cleanupWorkspace(t, researchWorkspace)
+
+	cfg := testCfg([]config.AgentConfig{
+		{ID: "main", Default: true, Workspace: mainWorkspace},
+		{ID: "research", Workspace: researchWorkspace},
+	})
+	cfg.Tools.ReadFile.Enabled = true
+	cfg.Tools.WriteFile.Enabled = true
+
+	registry := NewAgentRegistry(cfg, &mockRegistryProvider{})
+	mainAgent, ok := registry.GetAgent("main")
+	if !ok || mainAgent == nil {
+		t.Fatal("expected main agent")
+	}
+
+	messages := mainAgent.ContextBuilder.BuildMessages(
+		nil,
+		"",
+		"delegate wisely",
+		nil,
+		"telegram",
+		"chat-1",
+		"",
+		"",
+	)
+	if len(messages) == 0 {
+		t.Fatal("expected messages")
+	}
+
+	systemPrompt := messages[0].Content
+	if !strings.Contains(systemPrompt, "# Agent Discovery") {
+		t.Fatalf("expected discovery section in system prompt, got %q", systemPrompt)
+	}
+	if !strings.Contains(systemPrompt, `"current_agent_id": "main"`) {
+		t.Fatalf("expected current agent id in discovery section, got %q", systemPrompt)
+	}
+	if !strings.Contains(systemPrompt, `"id": "main"`) ||
+		!strings.Contains(systemPrompt, `"id": "research"`) {
+		t.Fatalf("expected self and peer descriptors in discovery section, got %q", systemPrompt)
+	}
+	if !strings.Contains(systemPrompt, `"available_tools": [`) ||
+		!strings.Contains(systemPrompt, `"read_file"`) ||
+		!strings.Contains(systemPrompt, `"write_file"`) {
+		t.Fatalf("expected visible tool list in discovery section, got %q", systemPrompt)
+	}
+}
+
+func TestContextBuilder_BuildMessagesOmitsAgentDiscoverySectionForSingleton(t *testing.T) {
+	mainWorkspace := setupWorkspace(t, map[string]string{
+		"AGENT.md": `---
+description: Main agent
+---
+# Agent
+
+Generalist.
+`,
+	})
+	defer cleanupWorkspace(t, mainWorkspace)
+
+	cfg := testCfg([]config.AgentConfig{
+		{ID: "main", Default: true, Workspace: mainWorkspace},
+	})
+	cfg.Tools.ReadFile.Enabled = true
+
+	registry := NewAgentRegistry(cfg, &mockRegistryProvider{})
+	mainAgent, ok := registry.GetAgent("main")
+	if !ok || mainAgent == nil {
+		t.Fatal("expected main agent")
+	}
+
+	messages := mainAgent.ContextBuilder.BuildMessages(
+		nil,
+		"",
+		"handle locally",
+		nil,
+		"telegram",
+		"chat-1",
+		"",
+		"",
+	)
+	if len(messages) == 0 {
+		t.Fatal("expected messages")
+	}
+
+	systemPrompt := messages[0].Content
+	if strings.Contains(systemPrompt, "# Agent Discovery") {
+		t.Fatalf("did not expect discovery section for singleton registry, got %q", systemPrompt)
+	}
+	if strings.Contains(systemPrompt, `"current_agent_id": "main"`) {
+		t.Fatalf("did not expect discovery payload for singleton registry, got %q", systemPrompt)
+	}
+}
diff --git a/pkg/agent/eventbus_test.go b/pkg/agent/eventbus_test.go
index 19a1ea9eb..0b0e351dd 100644
--- a/pkg/agent/eventbus_test.go
+++ b/pkg/agent/eventbus_test.go
@@ -275,7 +275,13 @@ func TestAgentLoop_EmitsSteeringAndSkippedToolEvents(t *testing.T) {
 
 	resultCh := make(chan string, 1)
 	go func() {
-		resp, _ := al.ProcessDirectWithChannel(context.Background(), "do something", "test-session", "test", "chat1")
+		resp, _ := al.ProcessDirectWithChannel(
+			context.Background(),
+			"do something",
+			"test-session",
+			"test",
+			"chat1",
+		)
 		resultCh <- resp
 	}()
 
@@ -338,7 +344,11 @@ func TestAgentLoop_EmitsSteeringAndSkippedToolEvents(t *testing.T) {
 		t.Fatalf("expected steering interrupt kind, got %q", interruptPayload.Kind)
 	}
 	if interruptPayload.ContentLen != len("change course") {
-		t.Fatalf("expected interrupt content len %d, got %d", len("change course"), interruptPayload.ContentLen)
+		t.Fatalf(
+			"expected interrupt content len %d, got %d",
+			len("change course"),
+			interruptPayload.ContentLen,
+		)
 	}
 }
 
@@ -360,7 +370,9 @@ func TestAgentLoop_EmitsContextCompressEventOnRetry(t *testing.T) {
 		},
 	}
 
-	contextErr := stringError("InvalidParameter: Total tokens of image and text exceed max message tokens")
+	contextErr := stringError(
+		"InvalidParameter: Total tokens of image and text exceed max message tokens",
+	)
 	provider := &failFirstMockProvider{
 		failures:    1,
 		failError:   contextErr,
@@ -603,7 +615,12 @@ func collectEventStream(ch <-chan Event) []Event {
 	}
 }
 
-func waitForEvent(t *testing.T, ch <-chan Event, timeout time.Duration, match func(Event) bool) Event {
+func waitForEvent(
+	t *testing.T,
+	ch <-chan Event,
+	timeout time.Duration,
+	match func(Event) bool,
+) Event {
 	t.Helper()
 
 	timer := time.NewTimer(timeout)
diff --git a/pkg/agent/hook_mount_test.go b/pkg/agent/hook_mount_test.go
index 85d8f5c11..068f8da10 100644
--- a/pkg/agent/hook_mount_test.go
+++ b/pkg/agent/hook_mount_test.go
@@ -40,7 +40,11 @@ func (h *builtinAutoHook) AfterLLM(
 	return next, HookDecision{Action: HookActionModify}, nil
 }
 
-func newConfiguredHookLoop(t *testing.T, provider *llmHookTestProvider, hooks config.HooksConfig) *AgentLoop {
+func newConfiguredHookLoop(
+	t *testing.T,
+	provider *llmHookTestProvider,
+	hooks config.HooksConfig,
+) *AgentLoop {
 	t.Helper()
 
 	cfg := &config.Config{
@@ -102,7 +106,13 @@ func TestAgentLoop_ProcessDirectWithChannel_AutoMountsBuiltinHook(t *testing.T)
 	})
 	defer al.Close()
 
-	resp, err := al.ProcessDirectWithChannel(context.Background(), "hello", "session-1", "cli", "direct")
+	resp, err := al.ProcessDirectWithChannel(
+		context.Background(),
+		"hello",
+		"session-1",
+		"cli",
+		"direct",
+	)
 	if err != nil {
 		t.Fatalf("ProcessDirectWithChannel failed: %v", err)
 	}
@@ -140,7 +150,13 @@ func TestAgentLoop_ProcessDirectWithChannel_AutoMountsProcessHook(t *testing.T)
 	})
 	defer al.Close()
 
-	resp, err := al.ProcessDirectWithChannel(context.Background(), "hello", "session-1", "cli", "direct")
+	resp, err := al.ProcessDirectWithChannel(
+		context.Background(),
+		"hello",
+		"session-1",
+		"cli",
+		"direct",
+	)
 	if err != nil {
 		t.Fatalf("ProcessDirectWithChannel failed: %v", err)
 	}
@@ -172,7 +188,13 @@ func TestAgentLoop_ProcessDirectWithChannel_InvalidConfiguredHookFails(t *testin
 	})
 	defer al.Close()
 
-	_, err := al.ProcessDirectWithChannel(context.Background(), "hello", "session-1", "cli", "direct")
+	_, err := al.ProcessDirectWithChannel(
+		context.Background(),
+		"hello",
+		"session-1",
+		"cli",
+		"direct",
+	)
 	if err == nil {
 		t.Fatal("expected invalid configured hook error")
 	}
diff --git a/pkg/agent/hook_process.go b/pkg/agent/hook_process.go
index e5632913d..9b623ce42 100644
--- a/pkg/agent/hook_process.go
+++ b/pkg/agent/hook_process.go
@@ -98,7 +98,11 @@ type processHookAfterToolResponse struct {
 	Result *ToolResultHookResponse `json:"result,omitempty"`
 }
 
-func NewProcessHook(ctx context.Context, name string, opts ProcessHookOptions) (*ProcessHook, error) {
+func NewProcessHook(
+	ctx context.Context,
+	name string,
+	opts ProcessHookOptions,
+) (*ProcessHook, error) {
 	if len(opts.Command) == 0 {
 		return nil, fmt.Errorf("process hook command is required")
 	}
@@ -262,7 +266,10 @@ func (ph *ProcessHook) AfterTool(
 	return resp.Result, HookDecision{Action: resp.Action, Reason: resp.Reason}, nil
 }
 
-func (ph *ProcessHook) ApproveTool(ctx context.Context, req *ToolApprovalRequest) (ApprovalDecision, error) {
+func (ph *ProcessHook) ApproveTool(
+	ctx context.Context,
+	req *ToolApprovalRequest,
+) (ApprovalDecision, error) {
 	if ph == nil || !ph.opts.ApproveTool {
 		return ApprovalDecision{Approved: true}, nil
 	}
@@ -473,7 +480,11 @@ func (ph *ProcessHook) removePending(id uint64) {
 	}
 }
 
-func (al *AgentLoop) MountProcessHook(ctx context.Context, name string, opts ProcessHookOptions) error {
+func (al *AgentLoop) MountProcessHook(
+	ctx context.Context,
+	name string,
+	opts ProcessHookOptions,
+) error {
 	if al == nil {
 		return fmt.Errorf("agent loop is nil")
 	}
diff --git a/pkg/agent/hooks.go b/pkg/agent/hooks.go
index c1ef58ffd..4f63d0652 100644
--- a/pkg/agent/hooks.go
+++ b/pkg/agent/hooks.go
@@ -79,8 +79,14 @@ type LLMInterceptor interface {
 }
 
 type ToolInterceptor interface {
-	BeforeTool(ctx context.Context, call *ToolCallHookRequest) (*ToolCallHookRequest, HookDecision, error)
-	AfterTool(ctx context.Context, result *ToolResultHookResponse) (*ToolResultHookResponse, HookDecision, error)
+	BeforeTool(
+		ctx context.Context,
+		call *ToolCallHookRequest,
+	) (*ToolCallHookRequest, HookDecision, error)
+	AfterTool(
+		ctx context.Context,
+		result *ToolResultHookResponse,
+	) (*ToolResultHookResponse, HookDecision, error)
 }
 
 type ToolApprover interface {
@@ -295,7 +301,10 @@ func (hm *HookManager) dispatchEvents() {
 	}
 }
 
-func (hm *HookManager) BeforeLLM(ctx context.Context, req *LLMHookRequest) (*LLMHookRequest, HookDecision) {
+func (hm *HookManager) BeforeLLM(
+	ctx context.Context,
+	req *LLMHookRequest,
+) (*LLMHookRequest, HookDecision) {
 	if hm == nil || req == nil {
 		return req, HookDecision{Action: HookActionContinue}
 	}
@@ -326,7 +335,10 @@ func (hm *HookManager) BeforeLLM(ctx context.Context, req *LLMHookRequest) (*LLM
 	return current, HookDecision{Action: HookActionContinue}
 }
 
-func (hm *HookManager) AfterLLM(ctx context.Context, resp *LLMHookResponse) (*LLMHookResponse, HookDecision) {
+func (hm *HookManager) AfterLLM(
+	ctx context.Context,
+	resp *LLMHookResponse,
+) (*LLMHookResponse, HookDecision) {
 	if hm == nil || resp == nil {
 		return resp, HookDecision{Action: HookActionContinue}
 	}
diff --git a/pkg/agent/hooks_test.go b/pkg/agent/hooks_test.go
index 49e1b1784..d112d4c07 100644
--- a/pkg/agent/hooks_test.go
+++ b/pkg/agent/hooks_test.go
@@ -293,7 +293,10 @@ func TestAgentLoop_Hooks_ToolInterceptorCanRewrite(t *testing.T) {
 
 type denyApprovalHook struct{}
 
-func (h *denyApprovalHook) ApproveTool(ctx context.Context, req *ToolApprovalRequest) (ApprovalDecision, error) {
+func (h *denyApprovalHook) ApproveTool(
+	ctx context.Context,
+	req *ToolApprovalRequest,
+) (ApprovalDecision, error) {
 	return ApprovalDecision{
 		Approved: false,
 		Reason:   "blocked",
diff --git a/pkg/agent/instance.go b/pkg/agent/instance.go
index 880725660..4b3b4b3ee 100644
--- a/pkg/agent/instance.go
+++ b/pkg/agent/instance.go
@@ -72,12 +72,16 @@ func NewAgentInstance(
 	// Compile path whitelist patterns from config.
 	allowReadPaths := buildAllowReadPatterns(cfg)
 	allowWritePaths := compilePatterns(cfg.Tools.AllowWritePaths)
+	agentToolAllowlist := resolveAgentToolAllowlist(agentCfg)
 
 	toolsRegistry := tools.NewToolRegistry()
+	toolsRegistry.SetAllowlist(agentToolAllowlist)
 
 	if cfg.Tools.IsToolEnabled("read_file") {
 		maxReadFileSize := cfg.Tools.ReadFile.MaxReadFileSize
-		toolsRegistry.Register(tools.NewReadFileTool(workspace, readRestrict, maxReadFileSize, allowReadPaths))
+		toolsRegistry.Register(
+			tools.NewReadFileTool(workspace, readRestrict, maxReadFileSize, allowReadPaths),
+		)
 	}
 	if cfg.Tools.IsToolEnabled("write_file") {
 		toolsRegistry.Register(tools.NewWriteFileTool(workspace, restrict, allowWritePaths))
@@ -180,8 +184,15 @@ func NewAgentInstance(
 		if len(resolved) > 0 {
 			lightModelCfg, err := resolvedModelConfig(cfg, rc.LightModel, workspace)
 			if err != nil {
-				logger.WarnCF("agent", "Routing light model config invalid; routing disabled",
-					map[string]any{"light_model": rc.LightModel, "agent_id": agentID, "error": err.Error()})
+				logger.WarnCF(
+					"agent",
+					"Routing light model config invalid; routing disabled",
+					map[string]any{
+						"light_model": rc.LightModel,
+						"agent_id":    agentID,
+						"error":       err.Error(),
+					},
+				)
 			} else {
 				lp, _, err := providers.CreateProviderFromConfig(lightModelCfg)
 				if err != nil {
@@ -234,7 +245,8 @@ func resolveAgentWorkspace(agentCfg *config.AgentConfig, defaults *config.AgentD
 		return expandHome(strings.TrimSpace(agentCfg.Workspace))
 	}
 	// Use the configured default workspace (respects PICOCLAW_HOME)
-	if agentCfg == nil || agentCfg.Default || agentCfg.ID == "" || routing.NormalizeAgentID(agentCfg.ID) == "main" {
+	if agentCfg == nil || agentCfg.Default || agentCfg.ID == "" ||
+		routing.NormalizeAgentID(agentCfg.ID) == "main" {
 		return expandHome(defaults.Workspace)
 	}
 	// For named agents without explicit workspace, use default workspace with agent ID suffix
diff --git a/pkg/agent/instance_test.go b/pkg/agent/instance_test.go
index e296a18cb..a933a6493 100644
--- a/pkg/agent/instance_test.go
+++ b/pkg/agent/instance_test.go
@@ -156,7 +156,11 @@ func TestNewAgentInstance_ResolveCandidatesFromModelListAlias(t *testing.T) {
 				t.Fatalf("len(Candidates) = %d, want 1", len(agent.Candidates))
 			}
 			if agent.Candidates[0].Provider != tt.wantProvider {
-				t.Fatalf("candidate provider = %q, want %q", agent.Candidates[0].Provider, tt.wantProvider)
+				t.Fatalf(
+					"candidate provider = %q, want %q",
+					agent.Candidates[0].Provider,
+					tt.wantProvider,
+				)
 			}
 			if agent.Candidates[0].Model != tt.wantModel {
 				t.Fatalf("candidate model = %q, want %q", agent.Candidates[0].Model, tt.wantModel)
diff --git a/pkg/agent/loop.go b/pkg/agent/loop.go
index ef2951365..2193bbad3 100644
--- a/pkg/agent/loop.go
+++ b/pkg/agent/loop.go
@@ -192,7 +192,11 @@ func registerSharedTools(
 				Proxy:                 cfg.Tools.Web.Proxy,
 			})
 			if err != nil {
-				logger.ErrorCF("agent", "Failed to create web search tool", map[string]any{"error": err.Error()})
+				logger.ErrorCF(
+					"agent",
+					"Failed to create web search tool",
+					map[string]any{"error": err.Error()},
+				)
 			} else if searchTool != nil {
 				agent.Tools.Register(searchTool)
 			}
@@ -205,7 +209,11 @@ func registerSharedTools(
 				cfg.Tools.Web.FetchLimitBytes,
 				cfg.Tools.Web.PrivateHostWhitelist)
 			if err != nil {
-				logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+				logger.ErrorCF(
+					"agent",
+					"Failed to create web fetch tool",
+					map[string]any{"error": err.Error()},
+				)
 			} else {
 				agent.Tools.Register(fetchTool)
 			}
@@ -475,7 +483,12 @@ func (al *AgentLoop) Run(ctx context.Context) error {
 							"queue_depth": al.pendingSteeringCountForScope(target.SessionKey),
 						})
 
-					continued, continueErr := al.Continue(ctx, target.SessionKey, target.Channel, target.ChatID)
+					continued, continueErr := al.Continue(
+						ctx,
+						target.SessionKey,
+						target.Channel,
+						target.ChatID,
+					)
 					if continueErr != nil {
 						logger.WarnCF("agent", "Failed to continue queued steering",
 							map[string]any{
@@ -503,14 +516,22 @@ func (al *AgentLoop) Run(ctx context.Context) error {
 							"queue_depth": al.pendingSteeringCountForScope(target.SessionKey),
 						})
 
-					continued, continueErr := al.Continue(ctx, target.SessionKey, target.Channel, target.ChatID)
+					continued, continueErr := al.Continue(
+						ctx,
+						target.SessionKey,
+						target.Channel,
+						target.ChatID,
+					)
 					if continueErr != nil {
-						logger.WarnCF("agent", "Failed to continue queued steering after shutdown drain",
+						logger.WarnCF(
+							"agent",
+							"Failed to continue queued steering after shutdown drain",
 							map[string]any{
 								"channel": target.Channel,
 								"chat_id": target.ChatID,
 								"error":   continueErr.Error(),
-							})
+							},
+						)
 						return
 					}
 					if continued == "" {
@@ -565,11 +586,15 @@ func (al *AgentLoop) drainBusToSteering(ctx context.Context, activeScope, active
 		msgScope, _, scopeOK := al.resolveSteeringTarget(msg)
 		if !scopeOK || msgScope != activeScope {
 			if err := al.requeueInboundMessage(msg); err != nil {
-				logger.WarnCF("agent", "Failed to requeue non-steering inbound message", map[string]any{
-					"error":     err.Error(),
-					"channel":   msg.Channel,
-					"sender_id": msg.SenderID,
-				})
+				logger.WarnCF(
+					"agent",
+					"Failed to requeue non-steering inbound message",
+					map[string]any{
+						"error":     err.Error(),
+						"channel":   msg.Channel,
+						"sender_id": msg.SenderID,
+					},
+				)
 			}
 			continue
 		}
@@ -603,7 +628,10 @@ func (al *AgentLoop) Stop() {
 	al.running.Store(false)
 }
 
-func (al *AgentLoop) PublishResponseIfNeeded(ctx context.Context, channel, chatID, response string) {
+func (al *AgentLoop) PublishResponseIfNeeded(
+	ctx context.Context,
+	channel, chatID, response string,
+) {
 	if response == "" {
 		return
 	}
@@ -1053,7 +1081,10 @@ var audioAnnotationRe = regexp.MustCompile(`\[(voice|audio)(?::[^\]]*)?\]`)
 // transcribeAudioInMessage resolves audio media refs, transcribes them, and
 // replaces audio annotations in msg.Content with the transcribed text.
 // Returns the (possibly modified) message and true if audio was transcribed.
-func (al *AgentLoop) transcribeAudioInMessage(ctx context.Context, msg bus.InboundMessage) (bus.InboundMessage, bool) {
+func (al *AgentLoop) transcribeAudioInMessage(
+	ctx context.Context,
+	msg bus.InboundMessage,
+) (bus.InboundMessage, bool) {
 	if al.transcriber == nil || al.mediaStore == nil || len(msg.Media) == 0 {
 		return msg, false
 	}
@@ -1063,7 +1094,11 @@ func (al *AgentLoop) transcribeAudioInMessage(ctx context.Context, msg bus.Inbou
 	for _, ref := range msg.Media {
 		path, meta, err := al.mediaStore.ResolveWithMeta(ref)
 		if err != nil {
-			logger.WarnCF("voice", "Failed to resolve media ref", map[string]any{"ref": ref, "error": err})
+			logger.WarnCF(
+				"voice",
+				"Failed to resolve media ref",
+				map[string]any{"ref": ref, "error": err},
+			)
 			continue
 		}
 		if !utils.IsAudioFile(meta.Filename, meta.ContentType) {
@@ -1141,7 +1176,11 @@ func (al *AgentLoop) sendTranscriptionFeedback(
 		ReplyToMessageID: messageID,
 	})
 	if err != nil {
-		logger.WarnCF("voice", "Failed to send transcription feedback", map[string]any{"error": err.Error()})
+		logger.WarnCF(
+			"voice",
+			"Failed to send transcription feedback",
+			map[string]any{"error": err.Error()},
+		)
 	}
 }
 
@@ -1342,7 +1381,9 @@ func (al *AgentLoop) processMessage(ctx context.Context, msg bus.InboundMessage)
 	return al.runAgentLoop(ctx, agent, opts)
 }
 
-func (al *AgentLoop) resolveMessageRoute(msg bus.InboundMessage) (routing.ResolvedRoute, *AgentInstance, error) {
+func (al *AgentLoop) resolveMessageRoute(
+	msg bus.InboundMessage,
+) (routing.ResolvedRoute, *AgentInstance, error) {
 	registry := al.GetRegistry()
 	route := registry.ResolveRoute(routing.RouteInput{
 		Channel:    msg.Channel,
@@ -1358,7 +1399,10 @@ func (al *AgentLoop) resolveMessageRoute(msg bus.InboundMessage) (routing.Resolv
 		agent = registry.GetDefaultAgent()
 	}
 	if agent == nil {
-		return routing.ResolvedRoute{}, nil, fmt.Errorf("no agent available for route (agent_id=%s)", route.AgentID)
+		return routing.ResolvedRoute{}, nil, fmt.Errorf(
+			"no agent available for route (agent_id=%s)",
+			route.AgentID,
+		)
 	}
 
 	return route, agent, nil
@@ -1683,7 +1727,11 @@ func (al *AgentLoop) runTurn(ctx context.Context, ts *turnState) (turnResult, er
 		ts.recordPersistedMessage(rootMsg)
 	}
 
-	activeCandidates, activeModel, usedLight := al.selectCandidates(ts.agent, ts.userMessage, messages)
+	activeCandidates, activeModel, usedLight := al.selectCandidates(
+		ts.agent,
+		ts.userMessage,
+		messages,
+	)
 	activeProvider := ts.agent.Provider
 	if usedLight && ts.agent.LightProvider != nil {
 		activeProvider = ts.agent.LightProvider
@@ -2656,12 +2704,15 @@ turnLoop:
 	}
 
 	if steerMsgs := al.dequeueSteeringMessagesForScope(ts.sessionKey); len(steerMsgs) > 0 {
-		logger.InfoCF("agent", "Steering arrived after turn completion; continuing turn before finalizing",
+		logger.InfoCF(
+			"agent",
+			"Steering arrived after turn completion; continuing turn before finalizing",
 			map[string]any{
 				"agent_id":       ts.agent.ID,
 				"steering_count": len(steerMsgs),
 				"session_key":    ts.sessionKey,
-			})
+			},
+		)
 		pendingMessages = append(pendingMessages, steerMsgs...)
 		finalContent = ""
 		goto turnLoop
@@ -2777,11 +2828,18 @@ func (al *AgentLoop) selectCandidates(
 			"score":       score,
 			"threshold":   agent.Router.Threshold(),
 		})
-	return agent.LightCandidates, resolvedCandidateModel(agent.LightCandidates, agent.Router.LightModel()), true
+	return agent.LightCandidates, resolvedCandidateModel(
+		agent.LightCandidates,
+		agent.Router.LightModel(),
+	), true
 }
 
 // maybeSummarize triggers summarization if the session history exceeds thresholds.
-func (al *AgentLoop) maybeSummarize(agent *AgentInstance, sessionKey string, turnScope turnEventScope) {
+func (al *AgentLoop) maybeSummarize(
+	agent *AgentInstance,
+	sessionKey string,
+	turnScope turnEventScope,
+) {
 	newHistory := agent.Sessions.GetHistory(sessionKey)
 	tokenEstimate := al.estimateTokens(newHistory)
 	threshold := agent.ContextWindow * agent.SummarizeTokenPercent / 100
@@ -2815,7 +2873,10 @@ type compressionResult struct {
 // prompt is built dynamically by BuildMessages and is NOT stored here.
 // The compression note is recorded in the session summary so that
 // BuildMessages can include it in the next system prompt.
-func (al *AgentLoop) forceCompression(agent *AgentInstance, sessionKey string) (compressionResult, bool) {
+func (al *AgentLoop) forceCompression(
+	agent *AgentInstance,
+	sessionKey string,
+) (compressionResult, bool) {
 	history := agent.Sessions.GetHistory(sessionKey)
 	if len(history) <= 2 {
 		return compressionResult{}, false
@@ -2968,7 +3029,11 @@ func formatToolsForLog(toolDefs []providers.ToolDefinition) string {
 }
 
 // summarizeSession summarizes the conversation history for a session.
-func (al *AgentLoop) summarizeSession(agent *AgentInstance, sessionKey string, turnScope turnEventScope) {
+func (al *AgentLoop) summarizeSession(
+	agent *AgentInstance,
+	sessionKey string,
+	turnScope turnEventScope,
+) {
 	ctx, cancel := context.WithTimeout(context.Background(), 120*time.Second)
 	defer cancel()
 
@@ -3320,7 +3385,10 @@ func (al *AgentLoop) applyExplicitSkillCommand(
 
 	skillName, ok := agent.ContextBuilder.ResolveSkillName(arg)
 	if !ok {
-		return true, true, fmt.Sprintf("Unknown skill: %s\nUse /list skills to see installed skills.", arg)
+		return true, true, fmt.Sprintf(
+			"Unknown skill: %s\nUse /list skills to see installed skills.",
+			arg,
+		)
 	}
 
 	if len(parts) < 3 {
@@ -3347,7 +3415,10 @@ func (al *AgentLoop) applyExplicitSkillCommand(
 	return true, false, ""
 }
 
-func (al *AgentLoop) buildCommandsRuntime(agent *AgentInstance, opts *processOptions) *commands.Runtime {
+func (al *AgentLoop) buildCommandsRuntime(
+	agent *AgentInstance,
+	opts *processOptions,
+) *commands.Runtime {
 	registry := al.GetRegistry()
 	cfg := al.GetConfig()
 	rt := &commands.Runtime{
@@ -3391,7 +3462,10 @@ func (al *AgentLoop) buildCommandsRuntime(agent *AgentInstance, opts *processOpt
 			rt.ListSkillNames = agent.ContextBuilder.ListSkillNames
 		}
 		rt.GetModelInfo = func() (string, string) {
-			return agent.Model, resolvedCandidateProvider(agent.Candidates, cfg.Agents.Defaults.Provider)
+			return agent.Model, resolvedCandidateProvider(
+				agent.Candidates,
+				cfg.Agents.Defaults.Provider,
+			)
 		}
 		rt.SwitchModel = func(value string) (string, error) {
 			value = strings.TrimSpace(value)
@@ -3405,7 +3479,12 @@ func (al *AgentLoop) buildCommandsRuntime(agent *AgentInstance, opts *processOpt
 				return "", fmt.Errorf("failed to initialize model %q: %w", value, err)
 			}
 
-			nextCandidates := resolveModelCandidates(cfg, cfg.Agents.Defaults.Provider, modelCfg.Model, agent.Fallbacks)
+			nextCandidates := resolveModelCandidates(
+				cfg,
+				cfg.Agents.Defaults.Provider,
+				modelCfg.Model,
+				agent.Fallbacks,
+			)
 			if len(nextCandidates) == 0 {
 				return "", fmt.Errorf("model %q did not resolve to any provider candidates", value)
 			}
diff --git a/pkg/agent/loop_mcp.go b/pkg/agent/loop_mcp.go
index 97debbc33..644f7168e 100644
--- a/pkg/agent/loop_mcp.go
+++ b/pkg/agent/loop_mcp.go
@@ -65,7 +65,11 @@ func (al *AgentLoop) ensureMCPInitialized(ctx context.Context) error {
 	}
 
 	if al.cfg.Tools.MCP.Servers == nil || len(al.cfg.Tools.MCP.Servers) == 0 {
-		logger.WarnCF("agent", "MCP is enabled but no servers are configured, skipping MCP initialization", nil)
+		logger.WarnCF(
+			"agent",
+			"MCP is enabled but no servers are configured, skipping MCP initialization",
+			nil,
+		)
 		return nil
 	}
 
@@ -76,7 +80,11 @@ func (al *AgentLoop) ensureMCPInitialized(ctx context.Context) error {
 		}
 	}
 	if !findValidServer {
-		logger.WarnCF("agent", "MCP is enabled but no valid servers are configured, skipping MCP initialization", nil)
+		logger.WarnCF(
+			"agent",
+			"MCP is enabled but no valid servers are configured, skipping MCP initialization",
+			nil,
+		)
 		return nil
 	}
 
@@ -193,10 +201,14 @@ func (al *AgentLoop) ensureMCPInitialized(ctx context.Context) error {
 				}
 
 				if useRegex {
-					agent.Tools.Register(tools.NewRegexSearchTool(agent.Tools, ttl, maxSearchResults))
+					agent.Tools.Register(
+						tools.NewRegexSearchTool(agent.Tools, ttl, maxSearchResults),
+					)
 				}
 				if useBM25 {
-					agent.Tools.Register(tools.NewBM25SearchTool(agent.Tools, ttl, maxSearchResults))
+					agent.Tools.Register(
+						tools.NewBM25SearchTool(agent.Tools, ttl, maxSearchResults),
+					)
 				}
 			}
 		}
diff --git a/pkg/agent/loop_media.go b/pkg/agent/loop_media.go
index e8314c10d..6958f51cb 100644
--- a/pkg/agent/loop_media.go
+++ b/pkg/agent/loop_media.go
@@ -25,7 +25,11 @@ import (
 // Non-image files (documents, audio, video) have their local path injected
 // into Content so the agent can access them via file tools like read_file.
 // Returns a new slice; original messages are not mutated.
-func resolveMediaRefs(messages []providers.Message, store media.MediaStore, maxSize int) []providers.Message {
+func resolveMediaRefs(
+	messages []providers.Message,
+	store media.MediaStore,
+	maxSize int,
+) []providers.Message {
 	if store == nil {
 		return messages
 	}
diff --git a/pkg/agent/loop_test.go b/pkg/agent/loop_test.go
index 25d20c689..9911c5cb7 100644
--- a/pkg/agent/loop_test.go
+++ b/pkg/agent/loop_test.go
@@ -591,7 +591,9 @@ func TestProcessMessage_MediaToolHandledSkipsFollowUpLLMAndFinalText(t *testing.
 	store := media.NewFileMediaStore()
 	al.SetMediaStore(store)
 	telegramChannel := &fakeMediaChannel{fakeChannel: fakeChannel{id: "rid-telegram"}}
-	al.SetChannelManager(newStartedTestChannelManager(t, msgBus, store, "telegram", telegramChannel))
+	al.SetChannelManager(
+		newStartedTestChannelManager(t, msgBus, store, "telegram", telegramChannel),
+	)
 
 	imagePath := filepath.Join(tmpDir, "screen.png")
 	if err := os.WriteFile(imagePath, []byte("fake screenshot"), 0o644); err != nil {
@@ -613,7 +615,10 @@ func TestProcessMessage_MediaToolHandledSkipsFollowUpLLMAndFinalText(t *testing.
 		t.Fatalf("processMessage() error = %v", err)
 	}
 	if response != "" {
-		t.Fatalf("expected no final response when media tool already handled delivery, got %q", response)
+		t.Fatalf(
+			"expected no final response when media tool already handled delivery, got %q",
+			response,
+		)
 	}
 	if provider.calls != 1 {
 		t.Fatalf("expected exactly 1 LLM call, got %d", provider.calls)
@@ -626,13 +631,20 @@ func TestProcessMessage_MediaToolHandledSkipsFollowUpLLMAndFinalText(t *testing.
 	}
 
 	if len(telegramChannel.sentMedia) != 1 {
-		t.Fatalf("expected exactly 1 synchronously sent media message, got %d", len(telegramChannel.sentMedia))
+		t.Fatalf(
+			"expected exactly 1 synchronously sent media message, got %d",
+			len(telegramChannel.sentMedia),
+		)
 	}
-	if telegramChannel.sentMedia[0].Channel != "telegram" || telegramChannel.sentMedia[0].ChatID != "chat1" {
+	if telegramChannel.sentMedia[0].Channel != "telegram" ||
+		telegramChannel.sentMedia[0].ChatID != "chat1" {
 		t.Fatalf("unexpected sent media target: %+v", telegramChannel.sentMedia[0])
 	}
 	if len(telegramChannel.sentMedia[0].Parts) != 1 {
-		t.Fatalf("expected exactly 1 sent media part, got %d", len(telegramChannel.sentMedia[0].Parts))
+		t.Fatalf(
+			"expected exactly 1 sent media part, got %d",
+			len(telegramChannel.sentMedia[0].Parts),
+		)
 	}
 
 	select {
@@ -660,7 +672,8 @@ func TestProcessMessage_MediaToolHandledSkipsFollowUpLLMAndFinalText(t *testing.
 		t.Fatal("expected session history to be saved")
 	}
 	last := history[len(history)-1]
-	if last.Role != "assistant" || last.Content != "Requested output delivered via tool attachment." {
+	if last.Role != "assistant" ||
+		last.Content != "Requested output delivered via tool attachment." {
 		t.Fatalf("expected handled assistant summary in history, got %+v", last)
 	}
 }
@@ -685,7 +698,9 @@ func TestProcessMessage_HandledToolProcessesQueuedSteeringBeforeReturning(t *tes
 	store := media.NewFileMediaStore()
 	al.SetMediaStore(store)
 	telegramChannel := &fakeMediaChannel{fakeChannel: fakeChannel{id: "rid-telegram"}}
-	al.SetChannelManager(newStartedTestChannelManager(t, msgBus, store, "telegram", telegramChannel))
+	al.SetChannelManager(
+		newStartedTestChannelManager(t, msgBus, store, "telegram", telegramChannel),
+	)
 
 	imagePath := filepath.Join(tmpDir, "screen-steering.png")
 	if err := os.WriteFile(imagePath, []byte("fake screenshot"), 0o644); err != nil {
@@ -714,7 +729,10 @@ func TestProcessMessage_HandledToolProcessesQueuedSteeringBeforeReturning(t *tes
 		t.Fatalf("expected 2 LLM calls after queued steering, got %d", provider.calls)
 	}
 	if len(telegramChannel.sentMedia) != 1 {
-		t.Fatalf("expected exactly 1 synchronously sent media message, got %d", len(telegramChannel.sentMedia))
+		t.Fatalf(
+			"expected exactly 1 synchronously sent media message, got %d",
+			len(telegramChannel.sentMedia),
+		)
 	}
 }
 
@@ -733,7 +751,9 @@ func TestProcessMessage_MediaArtifactCanBeForwardedBySendFile(t *testing.T) {
 	store := media.NewFileMediaStore()
 	al.SetMediaStore(store)
 	telegramChannel := &fakeMediaChannel{fakeChannel: fakeChannel{id: "rid-telegram"}}
-	al.SetChannelManager(newStartedTestChannelManager(t, msgBus, store, "telegram", telegramChannel))
+	al.SetChannelManager(
+		newStartedTestChannelManager(t, msgBus, store, "telegram", telegramChannel),
+	)
 
 	mediaDir := media.TempDir()
 	if err := os.MkdirAll(mediaDir, 0o700); err != nil {
@@ -766,13 +786,20 @@ func TestProcessMessage_MediaArtifactCanBeForwardedBySendFile(t *testing.T) {
 	}
 
 	if len(telegramChannel.sentMedia) != 1 {
-		t.Fatalf("expected exactly 1 synchronously sent media message, got %d", len(telegramChannel.sentMedia))
+		t.Fatalf(
+			"expected exactly 1 synchronously sent media message, got %d",
+			len(telegramChannel.sentMedia),
+		)
 	}
-	if telegramChannel.sentMedia[0].Channel != "telegram" || telegramChannel.sentMedia[0].ChatID != "chat1" {
+	if telegramChannel.sentMedia[0].Channel != "telegram" ||
+		telegramChannel.sentMedia[0].ChatID != "chat1" {
 		t.Fatalf("unexpected sent media target: %+v", telegramChannel.sentMedia[0])
 	}
 	if len(telegramChannel.sentMedia[0].Parts) != 1 {
-		t.Fatalf("expected exactly 1 sent media part, got %d", len(telegramChannel.sentMedia[0].Parts))
+		t.Fatalf(
+			"expected exactly 1 sent media part, got %d",
+			len(telegramChannel.sentMedia[0].Parts),
+		)
 	}
 
 	select {
@@ -1183,7 +1210,10 @@ func (m *handledMediaWithSteeringTool) Parameters() map[string]any {
 	}
 }
 
-func (m *handledMediaWithSteeringTool) Execute(ctx context.Context, args map[string]any) *tools.ToolResult {
+func (m *handledMediaWithSteeringTool) Execute(
+	ctx context.Context,
+	args map[string]any,
+) *tools.ToolResult {
 	if err := m.loop.Steer(providers.Message{Role: "user", Content: "what about this instead?"}); err != nil {
 		return tools.ErrorResult(err.Error()).WithError(err)
 	}
@@ -1336,7 +1366,11 @@ func newStrictChatCompletionTestServer(
 	}))
 }
 
-func (h testHelper) executeAndGetResponse(tb testing.TB, ctx context.Context, msg bus.InboundMessage) string {
+func (h testHelper) executeAndGetResponse(
+	tb testing.TB,
+	ctx context.Context,
+	msg bus.InboundMessage,
+) string {
 	// Use a short timeout to avoid hanging
 	timeoutCtx, cancel := context.WithTimeout(ctx, responseTimeout)
 	defer cancel()
@@ -1467,7 +1501,10 @@ func TestProcessMessage_CommandOutcomes(t *testing.T) {
 		t.Fatalf("unexpected /foo reply: %q", fooResp)
 	}
 	if provider.calls != 1 {
-		t.Fatalf("LLM should be called exactly once after /foo passthrough, calls=%d", provider.calls)
+		t.Fatalf(
+			"LLM should be called exactly once after /foo passthrough, calls=%d",
+			provider.calls,
+		)
 	}
 
 	newResp := helper.executeAndGetResponse(t, context.Background(), bus.InboundMessage{
@@ -1617,7 +1654,10 @@ func TestProcessMessage_SwitchModelRejectsUnknownAlias(t *testing.T) {
 	}
 
 	if provider.calls != 0 {
-		t.Fatalf("LLM should not be called for rejected /switch and /show, calls=%d", provider.calls)
+		t.Fatalf(
+			"LLM should not be called for rejected /switch and /show, calls=%d",
+			provider.calls,
+		)
 	}
 }
 
@@ -1635,7 +1675,13 @@ func TestProcessMessage_SwitchModelRoutesSubsequentRequestsToSelectedProvider(t
 
 	remoteCalls := 0
 	remoteModel := ""
-	remoteServer := newChatCompletionTestServer(t, "remote", "remote reply", &remoteCalls, &remoteModel)
+	remoteServer := newChatCompletionTestServer(
+		t,
+		"remote",
+		"remote reply",
+		&remoteCalls,
+		&remoteModel,
+	)
 	defer remoteServer.Close()
 
 	cfg := &config.Config{
@@ -1958,7 +2004,9 @@ func TestAgentLoop_ContextExhaustionRetry(t *testing.T) {
 	msgBus := bus.NewMessageBus()
 
 	// Create a provider that fails once with a context error
-	contextErr := fmt.Errorf("InvalidParameter: Total tokens of image and text exceed max message tokens")
+	contextErr := fmt.Errorf(
+		"InvalidParameter: Total tokens of image and text exceed max message tokens",
+	)
 	provider := &failFirstMockProvider{
 		failures:    1,
 		failError:   contextErr,
@@ -2039,7 +2087,13 @@ func TestAgentLoop_EmptyModelResponseUsesAccurateFallback(t *testing.T) {
 	provider := &simpleMockProvider{response: ""}
 	al := NewAgentLoop(cfg, msgBus, provider)
 
-	response, err := al.ProcessDirectWithChannel(context.Background(), "hello", "empty-response", "test", "chat1")
+	response, err := al.ProcessDirectWithChannel(
+		context.Background(),
+		"hello",
+		"empty-response",
+		"test",
+		"chat1",
+	)
 	if err != nil {
 		t.Fatalf("ProcessDirectWithChannel failed: %v", err)
 	}
@@ -2071,7 +2125,13 @@ func TestAgentLoop_ToolLimitUsesDedicatedFallback(t *testing.T) {
 	al := NewAgentLoop(cfg, msgBus, provider)
 	al.RegisterTool(&toolLimitTestTool{})
 
-	response, err := al.ProcessDirectWithChannel(context.Background(), "hello", "tool-limit", "test", "chat1")
+	response, err := al.ProcessDirectWithChannel(
+		context.Background(),
+		"hello",
+		"tool-limit",
+		"test",
+		"chat1",
+	)
 	if err != nil {
 		t.Fatalf("ProcessDirectWithChannel failed: %v", err)
 	}
@@ -2389,7 +2449,9 @@ func TestHandleReasoning(t *testing.T) {
 					break
 				}
 				if msg.Content == "should timeout" {
-					t.Fatal("expected reasoning message to be dropped when bus is full, but it was published")
+					t.Fatal(
+						"expected reasoning message to be dropped when bus is full, but it was published",
+					)
 				}
 			}
 		}
@@ -2483,7 +2545,12 @@ func TestProcessHeartbeat_DoesNotPublishToolFeedback(t *testing.T) {
 	provider := &toolFeedbackProvider{filePath: heartbeatFile}
 	al := NewAgentLoop(cfg, msgBus, provider)
 
-	response, err := al.ProcessHeartbeat(context.Background(), "check heartbeat tasks", "telegram", "chat-1")
+	response, err := al.ProcessHeartbeat(
+		context.Background(),
+		"check heartbeat tasks",
+		"telegram",
+		"chat-1",
+	)
 	if err != nil {
 		t.Fatalf("ProcessHeartbeat() error = %v", err)
 	}
@@ -2968,8 +3035,14 @@ func TestProcessMessage_ContextOverflowRecovery(t *testing.T) {
 	agent := al.GetRegistry().GetDefaultAgent()
 
 	for i := 0; i < 5; i++ {
-		agent.Sessions.AddFullMessage(sessionKey, providers.Message{Role: "user", Content: "heavy message"})
-		agent.Sessions.AddFullMessage(sessionKey, providers.Message{Role: "assistant", Content: "response"})
+		agent.Sessions.AddFullMessage(
+			sessionKey,
+			providers.Message{Role: "user", Content: "heavy message"},
+		)
+		agent.Sessions.AddFullMessage(
+			sessionKey,
+			providers.Message{Role: "assistant", Content: "response"},
+		)
 	}
 
 	response, err := al.processMessage(context.Background(), bus.InboundMessage{
diff --git a/pkg/agent/model_resolution.go b/pkg/agent/model_resolution.go
index 140cff718..d5c2f74ea 100644
--- a/pkg/agent/model_resolution.go
+++ b/pkg/agent/model_resolution.go
@@ -26,7 +26,8 @@ func buildModelListResolver(cfg *config.Config) func(raw string) (string, bool)
 			return "", false
 		}
 
-		if mc, err := cfg.GetModelConfig(raw); err == nil && mc != nil && strings.TrimSpace(mc.Model) != "" {
+		if mc, err := cfg.GetModelConfig(raw); err == nil && mc != nil &&
+			strings.TrimSpace(mc.Model) != "" {
 			return ensureProtocol(mc.Model), true
 		}
 
@@ -78,7 +79,10 @@ func resolvedCandidateProvider(candidates []providers.FallbackCandidate, fallbac
 	return fallback
 }
 
-func resolvedModelConfig(cfg *config.Config, modelName, workspace string) (*config.ModelConfig, error) {
+func resolvedModelConfig(
+	cfg *config.Config,
+	modelName, workspace string,
+) (*config.ModelConfig, error) {
 	if cfg == nil {
 		return nil, fmt.Errorf("config is nil")
 	}
diff --git a/pkg/agent/registry.go b/pkg/agent/registry.go
index 58b7ce440..46f54f5c8 100644
--- a/pkg/agent/registry.go
+++ b/pkg/agent/registry.go
@@ -12,6 +12,7 @@ import (
 
 // AgentRegistry manages multiple agent instances and routes messages to them.
 type AgentRegistry struct {
+	cfg      *config.Config
 	agents   map[string]*AgentInstance
 	resolver *routing.RouteResolver
 	mu       sync.RWMutex
@@ -23,6 +24,7 @@ func NewAgentRegistry(
 	provider providers.LLMProvider,
 ) *AgentRegistry {
 	registry := &AgentRegistry{
+		cfg:      cfg,
 		agents:   make(map[string]*AgentInstance),
 		resolver: routing.NewRouteResolver(cfg),
 	}
@@ -52,6 +54,14 @@ func NewAgentRegistry(
 		}
 	}
 
+	for id, instance := range registry.agents {
+		if instance.ContextBuilder != nil {
+			instance.ContextBuilder.
+				WithAgentIdentity(id).
+				WithAgentDiscovery(registry.ListAgents)
+		}
+	}
+
 	return registry
 }
 
@@ -130,11 +140,13 @@ func (r *AgentRegistry) Close() {
 func (r *AgentRegistry) GetDefaultAgent() *AgentInstance {
 	r.mu.RLock()
 	defer r.mu.RUnlock()
-	if agent, ok := r.agents["main"]; ok {
-		return agent
+	if id := r.defaultAgentIDLocked(); id != "" {
+		if agent, ok := r.agents[id]; ok {
+			return agent
+		}
 	}
-	for _, agent := range r.agents {
-		return agent
+	for id := range r.agents {
+		return r.agents[id]
 	}
 	return nil
 }
diff --git a/pkg/agent/registry_test.go b/pkg/agent/registry_test.go
index b173ef967..2b577ab93 100644
--- a/pkg/agent/registry_test.go
+++ b/pkg/agent/registry_test.go
@@ -2,8 +2,10 @@ package agent
 
 import (
 	"context"
+	"slices"
 	"testing"
 
+	"github.com/sipeed/picoclaw/pkg/bus"
 	"github.com/sipeed/picoclaw/pkg/config"
 	"github.com/sipeed/picoclaw/pkg/providers"
 )
@@ -200,6 +202,77 @@ func TestAgentInstance_FallbackExplicitEmpty(t *testing.T) {
 
 	agent, _ := registry.GetAgent("no-fallback")
 	if len(agent.Fallbacks) != 0 {
-		t.Errorf("expected 0 fallbacks (explicit empty), got %d: %v", len(agent.Fallbacks), agent.Fallbacks)
+		t.Errorf(
+			"expected 0 fallbacks (explicit empty), got %d: %v",
+			len(agent.Fallbacks),
+			agent.Fallbacks,
+		)
+	}
+}
+
+func TestNewAgentLoop_AgentToolAllowlistFiltersRuntimeTools(t *testing.T) {
+	cfg := testCfg([]config.AgentConfig{
+		{ID: "main", Default: true},
+		{
+			ID:    "research",
+			Tools: []string{"read_file", "write_file", "web_search", "web_fetch", "message"},
+		},
+	})
+	cfg.Tools.ReadFile.Enabled = true
+	cfg.Tools.WriteFile.Enabled = true
+	cfg.Tools.ListDir.Enabled = true
+	cfg.Tools.Exec.Enabled = true
+	cfg.Tools.Message.Enabled = true
+	cfg.Tools.Web.Enabled = true
+	cfg.Tools.Web.DuckDuckGo.Enabled = true
+	cfg.Tools.WebFetch.Enabled = true
+	cfg.Tools.Spawn.Enabled = true
+	cfg.Tools.Subagent.Enabled = true
+
+	al := NewAgentLoop(cfg, bus.NewMessageBus(), &mockRegistryProvider{})
+	defer al.Close()
+
+	research, ok := al.GetRegistry().GetAgent("research")
+	if !ok || research == nil {
+		t.Fatal("expected research agent")
+	}
+
+	got := research.Tools.List()
+	want := []string{"message", "read_file", "web_fetch", "web_search", "write_file"}
+	if !slices.Equal(got, want) {
+		t.Fatalf("research tools = %v, want %v", got, want)
+	}
+
+	for _, blocked := range []string{"exec", "list_dir", "spawn", "subagent"} {
+		if _, ok := research.Tools.Get(blocked); ok {
+			t.Fatalf("expected %q to be blocked by allowlist", blocked)
+		}
+	}
+}
+
+func TestNewAgentLoop_AgentToolAllowlistRequiresExactRuntimeToolNames(t *testing.T) {
+	cfg := testCfg([]config.AgentConfig{
+		{ID: "main", Default: true},
+		{
+			ID:    "research",
+			Tools: []string{"web"},
+		},
+	})
+	cfg.Tools.Web.Enabled = true
+	cfg.Tools.Web.DuckDuckGo.Enabled = true
+
+	al := NewAgentLoop(cfg, bus.NewMessageBus(), &mockRegistryProvider{})
+	defer al.Close()
+
+	research, ok := al.GetRegistry().GetAgent("research")
+	if !ok || research == nil {
+		t.Fatal("expected research agent")
+	}
+
+	if _, ok := research.Tools.Get("web_search"); ok {
+		t.Fatal("web_search should not be registered when allowlist contains only web")
+	}
+	if slices.Contains(research.Tools.List(), "web_search") {
+		t.Fatalf("research tools = %v, expected web_search to be absent", research.Tools.List())
 	}
 }
diff --git a/pkg/agent/steering.go b/pkg/agent/steering.go
index ad6613e8c..7ce918dd8 100644
--- a/pkg/agent/steering.go
+++ b/pkg/agent/steering.go
@@ -325,7 +325,10 @@ func (al *AgentLoop) agentForSession(sessionKey string) *AgentInstance {
 // user has since enqueued steering messages.
 //
 // If no steering messages are pending, it returns an empty string.
-func (al *AgentLoop) Continue(ctx context.Context, sessionKey, channel, chatID string) (string, error) {
+func (al *AgentLoop) Continue(
+	ctx context.Context,
+	sessionKey, channel, chatID string,
+) (string, error) {
 	if active := al.GetActiveTurn(); active != nil {
 		return "", fmt.Errorf("turn %s is still active", active.TurnID)
 	}
diff --git a/pkg/agent/steering_test.go b/pkg/agent/steering_test.go
index 75ba9861d..deb4f07c5 100644
--- a/pkg/agent/steering_test.go
+++ b/pkg/agent/steering_test.go
@@ -896,7 +896,10 @@ func TestAgentLoop_Run_AutoContinuesLateSteeringMessage(t *testing.T) {
 	defer cancelNoExtra()
 	select {
 	case out2 := <-msgBus.OutboundChan():
-		t.Fatalf("expected stale direct response to be suppressed, got extra outbound %q", out2.Content)
+		t.Fatalf(
+			"expected stale direct response to be suppressed, got extra outbound %q",
+			out2.Content,
+		)
 	case <-noExtraCtx.Done():
 	}
 
@@ -1044,7 +1047,11 @@ func TestAgentLoop_Continue_PreservesSteeringMedia(t *testing.T) {
 	if err = os.WriteFile(pngPath, pngHeader, 0o644); err != nil {
 		t.Fatalf("WriteFile failed: %v", err)
 	}
-	ref, err := store.Store(pngPath, media.MediaMeta{Filename: "steer.png", ContentType: "image/png"}, "test")
+	ref, err := store.Store(
+		pngPath,
+		media.MediaMeta{Filename: "steer.png", ContentType: "image/png"},
+		"test",
+	)
 	if err != nil {
 		t.Fatalf("Store failed: %v", err)
 	}
@@ -1236,7 +1243,10 @@ func TestAgentLoop_InterruptGraceful_UsesTerminalNoToolCall(t *testing.T) {
 		t.Fatalf("expected 2 provider calls, got %d", calls)
 	}
 	if terminalToolsCount != 0 {
-		t.Fatalf("expected graceful terminal call to disable tools, got %d tool defs", terminalToolsCount)
+		t.Fatalf(
+			"expected graceful terminal call to disable tools, got %d tool defs",
+			terminalToolsCount,
+		)
 	}
 
 	foundHint := false
@@ -1247,7 +1257,8 @@ func TestAgentLoop_InterruptGraceful_UsesTerminalNoToolCall(t *testing.T) {
 		if msg.Role == "user" && msg.Content == expectedHint {
 			foundHint = true
 		}
-		if msg.Role == "tool" && msg.ToolCallID == "call_2" && msg.Content == "Skipped due to graceful interrupt." {
+		if msg.Role == "tool" && msg.ToolCallID == "call_2" &&
+			msg.Content == "Skipped due to graceful interrupt." {
 			foundSkipped = true
 		}
 	}
@@ -1539,7 +1550,8 @@ func TestAgentLoop_Steering_SkippedToolsHaveErrorResults(t *testing.T) {
 
 	foundSkipped := false
 	for _, m := range msgs {
-		if m.Role == "tool" && m.ToolCallID == "call_2" && m.Content == "Skipped due to queued user message." {
+		if m.Role == "tool" && m.ToolCallID == "call_2" &&
+			m.Content == "Skipped due to queued user message." {
 			foundSkipped = true
 			break
 		}
@@ -1547,7 +1559,13 @@ func TestAgentLoop_Steering_SkippedToolsHaveErrorResults(t *testing.T) {
 	if !foundSkipped {
 		// Log what we actually got
 		for i, m := range msgs {
-			t.Logf("msg[%d]: role=%s toolCallID=%s content=%s", i, m.Role, m.ToolCallID, truncate(m.Content, 80))
+			t.Logf(
+				"msg[%d]: role=%s toolCallID=%s content=%s",
+				i,
+				m.Role,
+				m.ToolCallID,
+				truncate(m.Content, 80),
+			)
 		}
 		t.Fatal("expected skipped tool result for call_2")
 	}
diff --git a/pkg/agent/subturn.go b/pkg/agent/subturn.go
index f5ba412ab..4fcbb089c 100644
--- a/pkg/agent/subturn.go
+++ b/pkg/agent/subturn.go
@@ -505,7 +505,12 @@ func spawnSubTurn(
 // Event emissions:
 //   - SubTurnResultDeliveredEvent: successful delivery to channel
 //   - SubTurnOrphanResultEvent: delivery failed (parent finished or channel full)
-func deliverSubTurnResult(al *AgentLoop, parentTS *turnState, childID string, result *tools.ToolResult) {
+func deliverSubTurnResult(
+	al *AgentLoop,
+	parentTS *turnState,
+	childID string,
+	result *tools.ToolResult,
+) {
 	// Let GC clean up the pendingResults channel; parent Finish will no longer close it.
 	// We use defer/recover to catch any unlikely channel panics if it were ever closed.
 	defer func() {
@@ -516,9 +521,14 @@ func deliverSubTurnResult(al *AgentLoop, parentTS *turnState, childID string, re
 				"recover":   r,
 			})
 			if result != nil && al != nil {
-				al.emitEvent(EventKindSubTurnOrphan,
+				al.emitEvent(
+					EventKindSubTurnOrphan,
 					parentTS.eventMeta("deliverSubTurnResult", "subturn.orphan"),
-					SubTurnOrphanPayload{ParentTurnID: parentTS.turnID, ChildTurnID: childID, Reason: "panic"},
+					SubTurnOrphanPayload{
+						ParentTurnID: parentTS.turnID,
+						ChildTurnID:  childID,
+						Reason:       "panic",
+					},
 				)
 			}
 		}
@@ -531,9 +541,14 @@ func deliverSubTurnResult(al *AgentLoop, parentTS *turnState, childID string, re
 	// If parent turn has already finished, treat this as an orphan result
 	if isFinished || resultChan == nil {
 		if result != nil && al != nil {
-			al.emitEvent(EventKindSubTurnOrphan,
+			al.emitEvent(
+				EventKindSubTurnOrphan,
 				parentTS.eventMeta("deliverSubTurnResult", "subturn.orphan"),
-				SubTurnOrphanPayload{ParentTurnID: parentTS.turnID, ChildTurnID: childID, Reason: "parent_finished"},
+				SubTurnOrphanPayload{
+					ParentTurnID: parentTS.turnID,
+					ChildTurnID:  childID,
+					Reason:       "parent_finished",
+				},
 			)
 		}
 		return
diff --git a/pkg/agent/subturn_test.go b/pkg/agent/subturn_test.go
index 6a2ba835d..ef5a03b20 100644
--- a/pkg/agent/subturn_test.go
+++ b/pkg/agent/subturn_test.go
@@ -571,7 +571,8 @@ func TestHardAbortSessionRollback(t *testing.T) {
 	}
 
 	// Verify the content matches the initial state
-	if finalHistory[0].Content != "initial message 1" || finalHistory[1].Content != "initial response 1" {
+	if finalHistory[0].Content != "initial message 1" ||
+		finalHistory[1].Content != "initial response 1" {
 		t.Error("history content does not match initial state after rollback")
 	}
 }
@@ -1290,7 +1291,12 @@ func TestDeliverSubTurnResult_RaceWithFinish(t *testing.T) {
 	finalOrphan := orphanCount
 	mu.Unlock()
 
-	t.Logf("Delivered: %d, Orphan: %d, Total: %d", finalDelivered, finalOrphan, finalDelivered+finalOrphan)
+	t.Logf(
+		"Delivered: %d, Orphan: %d, Total: %d",
+		finalDelivered,
+		finalOrphan,
+		finalDelivered+finalOrphan,
+	)
 
 	// With the new drainPendingResults behavior, the total events may be >= numResults
 	// because Finish() drains remaining results from the channel and emits them as orphans.
diff --git a/pkg/agent/tool_allowlist.go b/pkg/agent/tool_allowlist.go
new file mode 100644
index 000000000..41b1fb98b
--- /dev/null
+++ b/pkg/agent/tool_allowlist.go
@@ -0,0 +1,30 @@
+package agent
+
+import (
+	"sort"
+	"strings"
+
+	"github.com/sipeed/picoclaw/pkg/config"
+)
+
+func resolveAgentToolAllowlist(agentCfg *config.AgentConfig) []string {
+	if agentCfg == nil || agentCfg.Tools == nil {
+		return nil
+	}
+
+	allowlist := make(map[string]struct{}, len(agentCfg.Tools))
+	for _, raw := range agentCfg.Tools {
+		trimmed := strings.ToLower(strings.TrimSpace(raw))
+		if trimmed == "" {
+			continue
+		}
+		allowlist[trimmed] = struct{}{}
+	}
+
+	result := make([]string, 0, len(allowlist))
+	for name := range allowlist {
+		result = append(result, name)
+	}
+	sort.Strings(result)
+	return result
+}
diff --git a/pkg/config/config.go b/pkg/config/config.go
index 533f45a44..8f793526b 100644
--- a/pkg/config/config.go
+++ b/pkg/config/config.go
@@ -106,18 +106,18 @@ const CurrentVersion = 1
 
 // Config is the current config structure with version support
 type Config struct {
-	Version   int             `json:"version"            yaml:"-"` // Config schema version for migration
-	Agents    AgentsConfig    `json:"agents"             yaml:"-"`
-	Bindings  []AgentBinding  `json:"bindings,omitempty" yaml:"-"`
-	Session   SessionConfig   `json:"session,omitempty"  yaml:"-"`
-	Channels  ChannelsConfig  `json:"channels"           yaml:"channels"`
-	ModelList SecureModelList `json:"model_list"         yaml:"model_list"` // New model-centric provider configuration
-	Gateway   GatewayConfig   `json:"gateway"            yaml:"-"`
-	Hooks     HooksConfig     `json:"hooks,omitempty"    yaml:"-"`
-	Tools     ToolsConfig     `json:"tools"              yaml:",inline"`
-	Heartbeat HeartbeatConfig `json:"heartbeat"          yaml:"-"`
-	Devices   DevicesConfig   `json:"devices"            yaml:"-"`
-	Voice     VoiceConfig     `json:"voice"              yaml:"-"`
+	Version   int             `json:"version"              yaml:"-"` // Config schema version for migration
+	Agents    AgentsConfig    `json:"agents"               yaml:"-"`
+	Bindings  []AgentBinding  `json:"bindings,omitempty"   yaml:"-"`
+	Session   SessionConfig   `json:"session,omitempty"    yaml:"-"`
+	Channels  ChannelsConfig  `json:"channels"             yaml:"channels"`
+	ModelList SecureModelList `json:"model_list"           yaml:"model_list"` // New model-centric provider configuration
+	Gateway   GatewayConfig   `json:"gateway"              yaml:"-"`
+	Hooks     HooksConfig     `json:"hooks,omitempty"      yaml:"-"`
+	Tools     ToolsConfig     `json:"tools"                yaml:",inline"`
+	Heartbeat HeartbeatConfig `json:"heartbeat"            yaml:"-"`
+	Devices   DevicesConfig   `json:"devices"              yaml:"-"`
+	Voice     VoiceConfig     `json:"voice"                yaml:"-"`
 	// BuildInfo contains build-time version information
 	BuildInfo BuildInfo `json:"build_info,omitempty" yaml:"-"`
 
@@ -248,6 +248,7 @@ type AgentConfig struct {
 	Name      string            `json:"name,omitempty"`
 	Workspace string            `json:"workspace,omitempty"`
 	Model     *AgentModelConfig `json:"model,omitempty"`
+	Tools     []string          `json:"tools,omitempty"`
 	Skills    []string          `json:"skills,omitempty"`
 	Subagents *SubagentsConfig  `json:"subagents,omitempty"`
 }
@@ -818,8 +819,8 @@ type GLMSearchConfig struct {
 	BaseURL string       `json:"base_url"         yaml:"-"                 env:"PICOCLAW_TOOLS_WEB_GLM_BASE_URL"`
 	// SearchEngine specifies the search backend: "search_std" (default),
 	// "search_pro", "search_pro_sogou", or "search_pro_quark".
-	SearchEngine string `json:"search_engine" yaml:"-" env:"PICOCLAW_TOOLS_WEB_GLM_SEARCH_ENGINE"`
-	MaxResults   int    `json:"max_results"   yaml:"-" env:"PICOCLAW_TOOLS_WEB_GLM_MAX_RESULTS"`
+	SearchEngine string `json:"search_engine"    yaml:"-"                 env:"PICOCLAW_TOOLS_WEB_GLM_SEARCH_ENGINE"`
+	MaxResults   int    `json:"max_results"      yaml:"-"                 env:"PICOCLAW_TOOLS_WEB_GLM_MAX_RESULTS"`
 }
 
 type BaiduSearchConfig struct {
@@ -830,7 +831,7 @@ type BaiduSearchConfig struct {
 }
 
 type WebToolsConfig struct {
-	ToolConfig  `                  yaml:"-"                      envPrefix:"PICOCLAW_TOOLS_WEB_"`
+	ToolConfig  `                    yaml:"-"                      envPrefix:"PICOCLAW_TOOLS_WEB_"`
 	Brave       BraveConfig       `yaml:"brave,omitempty"                                        json:"brave"`
 	Tavily      TavilyConfig      `yaml:"tavily,omitempty"                                       json:"tavily"`
 	DuckDuckGo  DuckDuckGoConfig  `yaml:"-"                                                      json:"duckduckgo"`
@@ -843,13 +844,13 @@ type WebToolsConfig struct {
 	// the client-side web_search tool is hidden to avoid duplicate search surfaces,
 	// and the provider's built-in search is used instead. Falls back to client-side
 	// search when the provider does not support native search.
-	PreferNative bool `json:"prefer_native" yaml:"-" env:"PICOCLAW_TOOLS_WEB_PREFER_NATIVE"`
+	PreferNative bool `yaml:"-"                                                      json:"prefer_native"                    env:"PICOCLAW_TOOLS_WEB_PREFER_NATIVE"`
 	// Proxy is an optional proxy URL for web tools (http/https/socks5/socks5h).
 	// For authenticated proxies, prefer HTTP_PROXY/HTTPS_PROXY env vars instead of embedding credentials in config.
-	Proxy                string              `json:"proxy,omitempty"                  yaml:"-" env:"PICOCLAW_TOOLS_WEB_PROXY"`
-	FetchLimitBytes      int64               `json:"fetch_limit_bytes,omitempty"      yaml:"-" env:"PICOCLAW_TOOLS_WEB_FETCH_LIMIT_BYTES"`
-	Format               string              `json:"format,omitempty"                 yaml:"-" env:"PICOCLAW_TOOLS_WEB_FORMAT"`
-	PrivateHostWhitelist FlexibleStringSlice `json:"private_host_whitelist,omitempty" yaml:"-" env:"PICOCLAW_TOOLS_WEB_PRIVATE_HOST_WHITELIST"`
+	Proxy                string              `yaml:"-"                                                      json:"proxy,omitempty"                  env:"PICOCLAW_TOOLS_WEB_PROXY"`
+	FetchLimitBytes      int64               `yaml:"-"                                                      json:"fetch_limit_bytes,omitempty"      env:"PICOCLAW_TOOLS_WEB_FETCH_LIMIT_BYTES"`
+	Format               string              `yaml:"-"                                                      json:"format,omitempty"                 env:"PICOCLAW_TOOLS_WEB_FORMAT"`
+	PrivateHostWhitelist FlexibleStringSlice `yaml:"-"                                                      json:"private_host_whitelist,omitempty" env:"PICOCLAW_TOOLS_WEB_PRIVATE_HOST_WHITELIST"`
 }
 
 type CronToolsConfig struct {
@@ -887,37 +888,37 @@ type ReadFileToolConfig struct {
 }
 
 type ToolsConfig struct {
-	AllowReadPaths  []string `json:"allow_read_paths"  yaml:"-" env:"PICOCLAW_TOOLS_ALLOW_READ_PATHS"`
-	AllowWritePaths []string `json:"allow_write_paths" yaml:"-" env:"PICOCLAW_TOOLS_ALLOW_WRITE_PATHS"`
+	AllowReadPaths  []string `json:"allow_read_paths"      yaml:"-"                env:"PICOCLAW_TOOLS_ALLOW_READ_PATHS"`
+	AllowWritePaths []string `json:"allow_write_paths"     yaml:"-"                env:"PICOCLAW_TOOLS_ALLOW_WRITE_PATHS"`
 	// FilterSensitiveData controls whether to filter sensitive values (API keys,
 	// tokens, secrets) from tool results before sending to the LLM.
 	// Default: true (enabled)
-	FilterSensitiveData bool `json:"filter_sensitive_data" yaml:"-" env:"PICOCLAW_TOOLS_FILTER_SENSITIVE_DATA"`
+	FilterSensitiveData bool `json:"filter_sensitive_data" yaml:"-"                env:"PICOCLAW_TOOLS_FILTER_SENSITIVE_DATA"`
 	// FilterMinLength is the minimum content length required for filtering.
 	// Content shorter than this will be returned unchanged for performance.
 	// Default: 8
-	FilterMinLength int                `json:"filter_min_length" yaml:"-"                env:"PICOCLAW_TOOLS_FILTER_MIN_LENGTH"`
-	Web             WebToolsConfig     `json:"web"               yaml:"web,omitempty"`
-	Cron            CronToolsConfig    `json:"cron"              yaml:"-"`
-	Exec            ExecConfig         `json:"exec"              yaml:"-"`
-	Skills          SkillsToolsConfig  `json:"skills"            yaml:"skills,omitempty"`
-	MediaCleanup    MediaCleanupConfig `json:"media_cleanup"     yaml:"-"`
-	MCP             MCPConfig          `json:"mcp"               yaml:"-"`
-	AppendFile      ToolConfig         `json:"append_file"       yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_APPEND_FILE_"`
-	EditFile        ToolConfig         `json:"edit_file"         yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_EDIT_FILE_"`
-	FindSkills      ToolConfig         `json:"find_skills"       yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_FIND_SKILLS_"`
-	I2C             ToolConfig         `json:"i2c"               yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_I2C_"`
-	InstallSkill    ToolConfig         `json:"install_skill"     yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_INSTALL_SKILL_"`
-	ListDir         ToolConfig         `json:"list_dir"          yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_LIST_DIR_"`
-	Message         ToolConfig         `json:"message"           yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_MESSAGE_"`
-	ReadFile        ReadFileToolConfig `json:"read_file"         yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_READ_FILE_"`
-	SendFile        ToolConfig         `json:"send_file"         yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_SEND_FILE_"`
-	Spawn           ToolConfig         `json:"spawn"             yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_SPAWN_"`
-	SpawnStatus     ToolConfig         `json:"spawn_status"      yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_SPAWN_STATUS_"`
-	SPI             ToolConfig         `json:"spi"               yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_SPI_"`
-	Subagent        ToolConfig         `json:"subagent"          yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_SUBAGENT_"`
-	WebFetch        ToolConfig         `json:"web_fetch"         yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_WEB_FETCH_"`
-	WriteFile       ToolConfig         `json:"write_file"        yaml:"-"                                                       envPrefix:"PICOCLAW_TOOLS_WRITE_FILE_"`
+	FilterMinLength int                `json:"filter_min_length"     yaml:"-"                env:"PICOCLAW_TOOLS_FILTER_MIN_LENGTH"`
+	Web             WebToolsConfig     `json:"web"                   yaml:"web,omitempty"`
+	Cron            CronToolsConfig    `json:"cron"                  yaml:"-"`
+	Exec            ExecConfig         `json:"exec"                  yaml:"-"`
+	Skills          SkillsToolsConfig  `json:"skills"                yaml:"skills,omitempty"`
+	MediaCleanup    MediaCleanupConfig `json:"media_cleanup"         yaml:"-"`
+	MCP             MCPConfig          `json:"mcp"                   yaml:"-"`
+	AppendFile      ToolConfig         `json:"append_file"           yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_APPEND_FILE_"`
+	EditFile        ToolConfig         `json:"edit_file"             yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_EDIT_FILE_"`
+	FindSkills      ToolConfig         `json:"find_skills"           yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_FIND_SKILLS_"`
+	I2C             ToolConfig         `json:"i2c"                   yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_I2C_"`
+	InstallSkill    ToolConfig         `json:"install_skill"         yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_INSTALL_SKILL_"`
+	ListDir         ToolConfig         `json:"list_dir"              yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_LIST_DIR_"`
+	Message         ToolConfig         `json:"message"               yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_MESSAGE_"`
+	ReadFile        ReadFileToolConfig `json:"read_file"             yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_READ_FILE_"`
+	SendFile        ToolConfig         `json:"send_file"             yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_SEND_FILE_"`
+	Spawn           ToolConfig         `json:"spawn"                 yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_SPAWN_"`
+	SpawnStatus     ToolConfig         `json:"spawn_status"          yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_SPAWN_STATUS_"`
+	SPI             ToolConfig         `json:"spi"                   yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_SPI_"`
+	Subagent        ToolConfig         `json:"subagent"              yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_SUBAGENT_"`
+	WebFetch        ToolConfig         `json:"web_fetch"             yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_WEB_FETCH_"`
+	WriteFile       ToolConfig         `json:"write_file"            yaml:"-"                                                           envPrefix:"PICOCLAW_TOOLS_WRITE_FILE_"`
 }
 
 // IsFilterSensitiveDataEnabled returns true if sensitive data filtering is enabled
@@ -985,10 +986,10 @@ type MCPServerConfig struct {
 
 // MCPConfig defines configuration for all MCP servers
 type MCPConfig struct {
-	ToolConfig `                    envPrefix:"PICOCLAW_TOOLS_MCP_"`
+	ToolConfig `                           envPrefix:"PICOCLAW_TOOLS_MCP_"`
 	Discovery  ToolDiscoveryConfig `                                json:"discovery"`
 	// Servers is a map of server name to server configuration
-	Servers map[string]MCPServerConfig `json:"servers,omitempty"`
+	Servers map[string]MCPServerConfig `                                json:"servers,omitempty"`
 }
 
 func LoadConfig(path string) (*Config, error) {
@@ -999,7 +1000,10 @@ func LoadConfig(path string) (*Config, error) {
 	data, err := os.ReadFile(path)
 	if err != nil {
 		if os.IsNotExist(err) {
-			logger.WarnF("config file not found, using default config", map[string]any{"path": path})
+			logger.WarnF(
+				"config file not found, using default config",
+				map[string]any{"path": path},
+			)
 			return DefaultConfig(), nil
 		}
 		logger.Errorf("failed to read config file: %v", err)
@@ -1022,7 +1026,10 @@ func LoadConfig(path string) (*Config, error) {
 	var cfg *Config
 	switch versionInfo.Version {
 	case 0:
-		logger.InfoF("config migrate start", map[string]any{"from": versionInfo.Version, "to": CurrentVersion})
+		logger.InfoF(
+			"config migrate start",
+			map[string]any{"from": versionInfo.Version, "to": CurrentVersion},
+		)
 		// Legacy config (no version field)
 		v, e := loadConfigV0(data)
 		if e != nil {
@@ -1030,10 +1037,16 @@ func LoadConfig(path string) (*Config, error) {
 		}
 		cfg, e = v.Migrate()
 		if e != nil {
-			logger.ErrorF("config migrate fail", map[string]any{"from": versionInfo.Version, "to": CurrentVersion})
+			logger.ErrorF(
+				"config migrate fail",
+				map[string]any{"from": versionInfo.Version, "to": CurrentVersion},
+			)
 			return nil, e
 		}
-		logger.InfoF("config migrate success", map[string]any{"from": versionInfo.Version, "to": CurrentVersion})
+		logger.InfoF(
+			"config migrate success",
+			map[string]any{"from": versionInfo.Version, "to": CurrentVersion},
+		)
 		err = makeBackup(path)
 		if err != nil {
 			return nil, err
@@ -1041,7 +1054,10 @@ func LoadConfig(path string) (*Config, error) {
 		// Load existing security config and merge with migrated one to prevent data loss
 		secErr := loadSecurityConfig(cfg, securityPath(path))
 		if secErr != nil && !os.IsNotExist(secErr) {
-			logger.WarnF("failed to load existing security config during migration", map[string]any{"error": secErr})
+			logger.WarnF(
+				"failed to load existing security config during migration",
+				map[string]any{"error": secErr},
+			)
 			return nil, fmt.Errorf("failed to load existing security config: %w", secErr)
 		}
 		defer func(cfg *Config) {
diff --git a/pkg/config/config_test.go b/pkg/config/config_test.go
index 75eb458b8..a22bcd7cb 100644
--- a/pkg/config/config_test.go
+++ b/pkg/config/config_test.go
@@ -113,17 +113,18 @@ func TestAgentConfig_FullParse(t *testing.T) {
 					"name": "Sales Bot",
 					"model": "gpt-4"
 				},
-				{
-					"id": "support",
-					"name": "Support Bot",
-					"model": {
-						"primary": "claude-opus",
-						"fallbacks": ["haiku"]
-					},
-					"subagents": {
-						"allow_agents": ["sales"]
-					}
+			{
+				"id": "support",
+				"name": "Support Bot",
+				"model": {
+					"primary": "claude-opus",
+					"fallbacks": ["haiku"]
+				},
+				"tools": ["read_file", "web_search"],
+				"subagents": {
+					"allow_agents": ["sales"]
 				}
+			}
 			]
 		},
 		"bindings": [
@@ -171,6 +172,10 @@ func TestAgentConfig_FullParse(t *testing.T) {
 	if len(support.Model.Fallbacks) != 1 || support.Model.Fallbacks[0] != "haiku" {
 		t.Errorf("support.Model.Fallbacks = %v", support.Model.Fallbacks)
 	}
+	if len(support.Tools) != 2 || support.Tools[0] != "read_file" ||
+		support.Tools[1] != "web_search" {
+		t.Errorf("support.Tools = %v", support.Tools)
+	}
 	if support.Subagents == nil || len(support.Subagents.AllowAgents) != 1 {
 		t.Errorf("support.Subagents = %+v", support.Subagents)
 	}
@@ -182,7 +187,8 @@ func TestAgentConfig_FullParse(t *testing.T) {
 	if binding.AgentID != "support" || binding.Match.Channel != "telegram" {
 		t.Errorf("binding = %+v", binding)
 	}
-	if binding.Match.Peer == nil || binding.Match.Peer.Kind != "direct" || binding.Match.Peer.ID != "user123" {
+	if binding.Match.Peer == nil || binding.Match.Peer.Kind != "direct" ||
+		binding.Match.Peer.ID != "user123" {
 		t.Errorf("binding.Match.Peer = %+v", binding.Match.Peer)
 	}
 
@@ -387,7 +393,9 @@ func TestSaveConfig_PreservesDisabledTelegramPlaceholder(t *testing.T) {
 		t.Fatalf("LoadConfig failed: %v", err)
 	}
 	if loaded.Channels.Telegram.Placeholder.Enabled {
-		t.Fatal("telegram placeholder should remain disabled after SaveConfig/LoadConfig round-trip")
+		t.Fatal(
+			"telegram placeholder should remain disabled after SaveConfig/LoadConfig round-trip",
+		)
 	}
 }
 
@@ -510,7 +518,9 @@ func TestLoadConfig_ToolFeedbackDefaultsFalseWhenUnset(t *testing.T) {
 		t.Fatalf("LoadConfig() error: %v", err)
 	}
 	if cfg.Agents.Defaults.ToolFeedback.Enabled {
-		t.Fatal("agents.defaults.tool_feedback.enabled should remain false when unset in config file")
+		t.Fatal(
+			"agents.defaults.tool_feedback.enabled should remain false when unset in config file",
+		)
 	}
 }
 
@@ -764,7 +774,10 @@ func TestDefaultConfig_SummarizationThresholds(t *testing.T) {
 	cfg := DefaultConfig()
 
 	if cfg.Agents.Defaults.SummarizeMessageThreshold != 20 {
-		t.Errorf("SummarizeMessageThreshold = %d, want 20", cfg.Agents.Defaults.SummarizeMessageThreshold)
+		t.Errorf(
+			"SummarizeMessageThreshold = %d, want 20",
+			cfg.Agents.Defaults.SummarizeMessageThreshold,
+		)
 	}
 	if cfg.Agents.Defaults.SummarizeTokenPercent != 75 {
 		t.Errorf("SummarizeTokenPercent = %d, want 75", cfg.Agents.Defaults.SummarizeTokenPercent)
@@ -806,7 +819,11 @@ func TestDefaultConfig_WorkspacePath_WithPicoclawHome(t *testing.T) {
 	want := filepath.Join("/custom/picoclaw/home", "workspace")
 
 	if cfg.Agents.Defaults.Workspace != want {
-		t.Errorf("Workspace path with PICOCLAW_HOME = %q, want %q", cfg.Agents.Defaults.Workspace, want)
+		t.Errorf(
+			"Workspace path with PICOCLAW_HOME = %q, want %q",
+			cfg.Agents.Defaults.Workspace,
+			want,
+		)
 	}
 }
 
@@ -885,7 +902,12 @@ func TestFlexibleStringSlice_UnmarshalText(t *testing.T) {
 			}
 
 			if len(f) != len(tt.expected) {
-				t.Errorf("UnmarshalText(%q) length = %d, want %d", tt.input, len(f), len(tt.expected))
+				t.Errorf(
+					"UnmarshalText(%q) length = %d, want %d",
+					tt.input,
+					len(f),
+					len(tt.expected),
+				)
 				return
 			}
 
@@ -1006,7 +1028,8 @@ func TestLoadConfig_TelegramPlaceholderTextAcceptsSingleString(t *testing.T) {
 	if err != nil {
 		t.Fatalf("LoadConfig() error = %v", err)
 	}
-	if got := []string(cfg.Channels.Telegram.Placeholder.Text); len(got) != 1 || got[0] != "Thinking..." {
+	if got := []string(cfg.Channels.Telegram.Placeholder.Text); len(got) != 1 ||
+		got[0] != "Thinking..." {
 		t.Fatalf("placeholder.text = %#v, want [\"Thinking...\"]", got)
 	}
 }
@@ -1196,9 +1219,21 @@ func TestSaveConfig_MixedKeys(t *testing.T) {
 	cfg := &Config{
 		Version: CurrentVersion,
 		ModelList: []*ModelConfig{
-			{ModelName: "plain", Model: "openai/gpt-4", APIKeys: SimpleSecureStrings("sk-new-plaintext")},
-			{ModelName: "enc", Model: "openai/gpt-4", APIKeys: SimpleSecureStrings(alreadyEncrypted)},
-			{ModelName: "file", Model: "openai/gpt-4", APIKeys: SimpleSecureStrings("file://api.key")},
+			{
+				ModelName: "plain",
+				Model:     "openai/gpt-4",
+				APIKeys:   SimpleSecureStrings("sk-new-plaintext"),
+			},
+			{
+				ModelName: "enc",
+				Model:     "openai/gpt-4",
+				APIKeys:   SimpleSecureStrings(alreadyEncrypted),
+			},
+			{
+				ModelName: "file",
+				Model:     "openai/gpt-4",
+				APIKeys:   SimpleSecureStrings("file://api.key"),
+			},
 		},
 	}
 	if err := SaveConfig(cfgPath, cfg); err != nil {
@@ -1335,7 +1370,10 @@ func TestSaveConfig_UsesPassphraseProvider(t *testing.T) {
 
 	raw, _ := os.ReadFile(filepath.Join(dir, SecurityConfigFile))
 	if !strings.Contains(string(raw), "enc://") {
-		t.Errorf("SaveConfig should have encrypted plaintext key via PassphraseProvider; got:\n%s", raw)
+		t.Errorf(
+			"SaveConfig should have encrypted plaintext key via PassphraseProvider; got:\n%s",
+			raw,
+		)
 	}
 }
 
@@ -1587,9 +1625,13 @@ func TestFilterSensitiveData_AllTokenTypes(t *testing.T) {
 			FilterMinLength:     8,
 			// Web tool API keys
 			Web: WebToolsConfig{
-				Brave:       BraveConfig{APIKeys: SecureStrings{NewSecureString("brave-api-key")}},
-				Tavily:      TavilyConfig{APIKeys: SecureStrings{NewSecureString("tavily-api-key")}},
-				Perplexity:  PerplexityConfig{APIKeys: SecureStrings{NewSecureString("perplexity-api-key")}},
+				Brave: BraveConfig{APIKeys: SecureStrings{NewSecureString("brave-api-key")}},
+				Tavily: TavilyConfig{
+					APIKeys: SecureStrings{NewSecureString("tavily-api-key")},
+				},
+				Perplexity: PerplexityConfig{
+					APIKeys: SecureStrings{NewSecureString("perplexity-api-key")},
+				},
 				GLMSearch:   GLMSearchConfig{APIKey: *NewSecureString("glm-search-key")},
 				BaiduSearch: BaiduSearchConfig{APIKey: *NewSecureString("baidu-search-key")},
 			},
@@ -1597,7 +1639,9 @@ func TestFilterSensitiveData_AllTokenTypes(t *testing.T) {
 			Skills: SkillsToolsConfig{
 				Github: SkillsGithubConfig{Token: *NewSecureString("github-token-xyz")},
 				Registries: SkillsRegistriesConfig{
-					ClawHub: ClawHubRegistryConfig{AuthToken: *NewSecureString("clawhub-auth-token")},
+					ClawHub: ClawHubRegistryConfig{
+						AuthToken: *NewSecureString("clawhub-auth-token"),
+					},
 				},
 			},
 		},
diff --git a/pkg/config/defaults.go b/pkg/config/defaults.go
index bc4ab0649..20e2e531d 100644
--- a/pkg/config/defaults.go
+++ b/pkg/config/defaults.go
@@ -65,7 +65,11 @@ func DefaultConfig() *Config {
 					Enabled: true,
 					Text:    FlexibleStringSlice{"Thinking... 💭"},
 				},
-				Streaming:     StreamingConfig{Enabled: true, ThrottleSeconds: 3, MinGrowthChars: 200},
+				Streaming: StreamingConfig{
+					Enabled:         true,
+					ThrottleSeconds: 3,
+					MinGrowthChars:  200,
+				},
 				UseMarkdownV2: false,
 			},
 			Feishu: FeishuConfig{
diff --git a/pkg/config/migration.go b/pkg/config/migration.go
index fee800a76..43f9645a2 100644
--- a/pkg/config/migration.go
+++ b/pkg/config/migration.go
@@ -335,7 +335,8 @@ func v0ConvertProvidersToModelList(cfg *configV0) []modelConfigV0 {
 			providerNames: []string{"github_copilot", "copilot"},
 			protocol:      "github-copilot",
 			buildConfig: func(p providersConfigV0) (modelConfigV0, bool) {
-				if p.GitHubCopilot.APIKey == "" && p.GitHubCopilot.APIBase == "" && p.GitHubCopilot.ConnectMode == "" {
+				if p.GitHubCopilot.APIKey == "" && p.GitHubCopilot.APIBase == "" &&
+					p.GitHubCopilot.ConnectMode == "" {
 					return modelConfigV0{}, false
 				}
 				return modelConfigV0{
diff --git a/pkg/config/migration_integration_test.go b/pkg/config/migration_integration_test.go
index bc8160967..b6a70c2ef 100644
--- a/pkg/config/migration_integration_test.go
+++ b/pkg/config/migration_integration_test.go
@@ -72,7 +72,11 @@ func TestMigration_Integration_LegacyConfigWithoutWorkspace(t *testing.T) {
 	// CRITICAL: Verify that user's settings are preserved
 	// This was the bug - these settings were lost when Workspace was empty
 	if cfg.Agents.Defaults.Provider != "openai" {
-		t.Errorf("Provider = %q, want %q (user's setting should be preserved)", cfg.Agents.Defaults.Provider, "openai")
+		t.Errorf(
+			"Provider = %q, want %q (user's setting should be preserved)",
+			cfg.Agents.Defaults.Provider,
+			"openai",
+		)
 	}
 	// Old "model" field is migrated to "model_name" field
 	if cfg.Agents.Defaults.ModelName != "gpt-4o" {
@@ -299,7 +303,11 @@ func TestMigration_Integration_PreservesAllAgentsFields(t *testing.T) {
 		t.Errorf("Agent.ID = %q, want %q", cfg.Agents.List[0].ID, "special-agent")
 	}
 	if cfg.Agents.List[0].Workspace != "/special/workspace" {
-		t.Errorf("Agent.Workspace = %q, want %q", cfg.Agents.List[0].Workspace, "/special/workspace")
+		t.Errorf(
+			"Agent.Workspace = %q, want %q",
+			cfg.Agents.List[0].Workspace,
+			"/special/workspace",
+		)
 	}
 
 	// Workspace should have default since it was empty in legacy config
@@ -362,7 +370,10 @@ func TestMigration_Integration_ChannelsConfigMigrated(t *testing.T) {
 
 	// OneBot: group_trigger_prefix should be migrated to group_trigger.prefixes
 	if len(cfg.Channels.OneBot.GroupTrigger.Prefixes) != 2 {
-		t.Errorf("len(OneBot.GroupTrigger.Prefixes) = %d, want 2", len(cfg.Channels.OneBot.GroupTrigger.Prefixes))
+		t.Errorf(
+			"len(OneBot.GroupTrigger.Prefixes) = %d, want 2",
+			len(cfg.Channels.OneBot.GroupTrigger.Prefixes),
+		)
 	} else {
 		if cfg.Channels.OneBot.GroupTrigger.Prefixes[0] != "/" {
 			t.Errorf("Prefixes[0] = %q, want %q", cfg.Channels.OneBot.GroupTrigger.Prefixes[0], "/")
@@ -443,13 +454,25 @@ func TestMigration_Integration_RoundTrip_SerializeAndLoad(t *testing.T) {
 
 	// Verify configs are identical
 	if cfg2.Agents.Defaults.Provider != cfg1.Agents.Defaults.Provider {
-		t.Errorf("Provider changed from %q to %q", cfg1.Agents.Defaults.Provider, cfg2.Agents.Defaults.Provider)
+		t.Errorf(
+			"Provider changed from %q to %q",
+			cfg1.Agents.Defaults.Provider,
+			cfg2.Agents.Defaults.Provider,
+		)
 	}
 	if cfg2.Agents.Defaults.ModelName != cfg1.Agents.Defaults.ModelName {
-		t.Errorf("ModelName changed from %q to %q", cfg1.Agents.Defaults.ModelName, cfg2.Agents.Defaults.ModelName)
+		t.Errorf(
+			"ModelName changed from %q to %q",
+			cfg1.Agents.Defaults.ModelName,
+			cfg2.Agents.Defaults.ModelName,
+		)
 	}
 	if cfg2.Agents.Defaults.MaxTokens != cfg1.Agents.Defaults.MaxTokens {
-		t.Errorf("MaxTokens changed from %d to %d", cfg1.Agents.Defaults.MaxTokens, cfg2.Agents.Defaults.MaxTokens)
+		t.Errorf(
+			"MaxTokens changed from %d to %d",
+			cfg1.Agents.Defaults.MaxTokens,
+			cfg2.Agents.Defaults.MaxTokens,
+		)
 	}
 }
 
@@ -557,7 +580,11 @@ func TestMigration_Integration_ModelNameField(t *testing.T) {
 
 	// GetModelName() should return model_name, not model (deprecated)
 	if cfg.Agents.Defaults.GetModelName() != "deepseek-reasoner" {
-		t.Errorf("GetModelName() = %q, want %q", cfg.Agents.Defaults.GetModelName(), "deepseek-reasoner")
+		t.Errorf(
+			"GetModelName() = %q, want %q",
+			cfg.Agents.Defaults.GetModelName(),
+			"deepseek-reasoner",
+		)
 	}
 
 	if len(cfg.Agents.Defaults.ModelFallbacks) != 1 {
diff --git a/pkg/config/migration_test.go b/pkg/config/migration_test.go
index aeabe9730..1ae3c7b71 100644
--- a/pkg/config/migration_test.go
+++ b/pkg/config/migration_test.go
@@ -91,9 +91,11 @@ func TestConvertProvidersToModelList_LiteLLM(t *testing.T) {
 func TestConvertProvidersToModelList_Multiple(t *testing.T) {
 	cfg := &configV0{
 		Providers: providersConfigV0{
-			OpenAI: openAIProviderConfigV0{providerConfigV0: providerConfigV0{APIKey: "openai-key"}},
-			Groq:   providerConfigV0{APIKey: "groq-key"},
-			Zhipu:  providerConfigV0{APIKey: "zhipu-key"},
+			OpenAI: openAIProviderConfigV0{
+				providerConfigV0: providerConfigV0{APIKey: "openai-key"},
+			},
+			Groq:  providerConfigV0{APIKey: "groq-key"},
+			Zhipu: providerConfigV0{APIKey: "zhipu-key"},
 		},
 	}
 
@@ -142,8 +144,13 @@ func TestConvertProvidersToModelList_AllProviders(t *testing.T) {
 	// Other providers have no configuration, so they won't be converted.
 	cfg := &configV0{
 		Providers: providersConfigV0{
-			OpenAI:        openAIProviderConfigV0{providerConfigV0: providerConfigV0{APIKey: "key1"}},
-			LiteLLM:       providerConfigV0{APIKey: "key-litellm", APIBase: "http://localhost:4000/v1"},
+			OpenAI: openAIProviderConfigV0{
+				providerConfigV0: providerConfigV0{APIKey: "key1"},
+			},
+			LiteLLM: providerConfigV0{
+				APIKey:  "key-litellm",
+				APIBase: "http://localhost:4000/v1",
+			},
 			Anthropic:     providerConfigV0{APIKey: "key2"},
 			OpenRouter:    providerConfigV0{APIKey: "key3"},
 			Groq:          providerConfigV0{APIKey: "key4"},
@@ -261,7 +268,11 @@ func TestConvertProvidersToModelList_PreservesUserModel_DeepSeek(t *testing.T) {
 
 	// Should use user's model, not default
 	if result[0].Model != "deepseek/deepseek-reasoner" {
-		t.Errorf("Model = %q, want %q (user's configured model)", result[0].Model, "deepseek/deepseek-reasoner")
+		t.Errorf(
+			"Model = %q, want %q (user's configured model)",
+			result[0].Model,
+			"deepseek/deepseek-reasoner",
+		)
 	}
 }
 
@@ -371,7 +382,9 @@ func TestConvertProvidersToModelList_MultipleProviders_PreservesUserModel(t *tes
 			},
 		},
 		Providers: providersConfigV0{
-			OpenAI:   openAIProviderConfigV0{providerConfigV0: providerConfigV0{APIKey: "sk-openai"}},
+			OpenAI: openAIProviderConfigV0{
+				providerConfigV0: providerConfigV0{APIKey: "sk-openai"},
+			},
 			DeepSeek: providerConfigV0{APIKey: "sk-deepseek"},
 		},
 	}
@@ -391,7 +404,11 @@ func TestConvertProvidersToModelList_MultipleProviders_PreservesUserModel(t *tes
 			}
 		case "deepseek":
 			if mc.Model != "deepseek/deepseek-reasoner" {
-				t.Errorf("DeepSeek Model = %q, want %q (user's)", mc.Model, "deepseek/deepseek-reasoner")
+				t.Errorf(
+					"DeepSeek Model = %q, want %q (user's)",
+					mc.Model,
+					"deepseek/deepseek-reasoner",
+				)
 			}
 		}
 	}
@@ -489,7 +506,11 @@ func TestConvertProvidersToModelList_NoProviderField_SingleProvider(t *testing.T
 
 	// ModelName should be the user's model value for backward compatibility
 	if result[0].ModelName != "glm-4.7" {
-		t.Errorf("ModelName = %q, want %q (user's model for backward compatibility)", result[0].ModelName, "glm-4.7")
+		t.Errorf(
+			"ModelName = %q, want %q (user's model for backward compatibility)",
+			result[0].ModelName,
+			"glm-4.7",
+		)
 	}
 
 	// Model should use the user's model with protocol prefix
@@ -510,8 +531,10 @@ func TestConvertProvidersToModelList_NoProviderField_MultipleProviders(t *testin
 			},
 		},
 		Providers: providersConfigV0{
-			OpenAI: openAIProviderConfigV0{providerConfigV0: providerConfigV0{APIKey: "openai-key"}},
-			Zhipu:  providerConfigV0{APIKey: "zhipu-key"},
+			OpenAI: openAIProviderConfigV0{
+				providerConfigV0: providerConfigV0{APIKey: "openai-key"},
+			},
+			Zhipu: providerConfigV0{APIKey: "zhipu-key"},
 		},
 	}
 
@@ -571,7 +594,11 @@ func TestBuildModelWithProtocol_NoPrefix(t *testing.T) {
 func TestBuildModelWithProtocol_AlreadyHasPrefix(t *testing.T) {
 	result := buildModelWithProtocol("openrouter", "openrouter/auto")
 	if result != "openrouter/auto" {
-		t.Errorf("buildModelWithProtocol(openrouter, openrouter/auto) = %q, want %q", result, "openrouter/auto")
+		t.Errorf(
+			"buildModelWithProtocol(openrouter, openrouter/auto) = %q, want %q",
+			result,
+			"openrouter/auto",
+		)
 	}
 }
 
@@ -613,6 +640,10 @@ func TestConvertProvidersToModelList_LegacyModelWithProtocolPrefix(t *testing.T)
 
 	// Model should NOT have duplicated prefix
 	if result[0].Model != "openrouter/auto" {
-		t.Errorf("Model = %q, want %q (should not duplicate prefix)", result[0].Model, "openrouter/auto")
+		t.Errorf(
+			"Model = %q, want %q (should not duplicate prefix)",
+			result[0].Model,
+			"openrouter/auto",
+		)
 	}
 }
diff --git a/pkg/config/model_config_test.go b/pkg/config/model_config_test.go
index 6e88f4783..f001885af 100644
--- a/pkg/config/model_config_test.go
+++ b/pkg/config/model_config_test.go
@@ -17,7 +17,11 @@ func TestGetModelConfig_Found(t *testing.T) {
 		Version: CurrentVersion,
 		ModelList: []*ModelConfig{
 			{ModelName: "test-model", Model: "openai/gpt-4o", APIKeys: SimpleSecureStrings("key1")},
-			{ModelName: "other-model", Model: "anthropic/claude", APIKeys: SimpleSecureStrings("key2")},
+			{
+				ModelName: "other-model",
+				Model:     "anthropic/claude",
+				APIKeys:   SimpleSecureStrings("key2"),
+			},
 		},
 	}
 
@@ -114,8 +118,16 @@ func TestGetModelConfig_RoundRobinStartsFromFirstMatch(t *testing.T) {
 func TestGetModelConfig_Concurrent(t *testing.T) {
 	cfg := &Config{
 		ModelList: []*ModelConfig{
-			{ModelName: "concurrent-model", Model: "openai/gpt-4o-1", APIKeys: SimpleSecureStrings("key1")},
-			{ModelName: "concurrent-model", Model: "openai/gpt-4o-2", APIKeys: SimpleSecureStrings("key2")},
+			{
+				ModelName: "concurrent-model",
+				Model:     "openai/gpt-4o-1",
+				APIKeys:   SimpleSecureStrings("key1"),
+			},
+			{
+				ModelName: "concurrent-model",
+				Model:     "openai/gpt-4o-2",
+				APIKeys:   SimpleSecureStrings("key2"),
+			},
 		},
 	}
 
@@ -290,7 +302,11 @@ func TestConfig_ValidateModelList(t *testing.T) {
 			}
 			if err != nil && tt.errMsg != "" {
 				if !strings.Contains(err.Error(), tt.errMsg) {
-					t.Errorf("ValidateModelList() error = %v, want error containing %q", err, tt.errMsg)
+					t.Errorf(
+						"ValidateModelList() error = %v, want error containing %q",
+						err,
+						tt.errMsg,
+					)
 				}
 			}
 		})
diff --git a/pkg/config/multikey_test.go b/pkg/config/multikey_test.go
index e58c6dc9e..28fd9ff7d 100644
--- a/pkg/config/multikey_test.go
+++ b/pkg/config/multikey_test.go
@@ -117,7 +117,10 @@ func TestExpandMultiKeyModels_WithExistingFallbacks(t *testing.T) {
 		ModelName: "gpt-4",
 		Model:     "openai/gpt-4o",
 	}
-	modelCfg.APIKeys = SimpleSecureStrings("key0", "key1") // Use internal field for multi-key testing
+	modelCfg.APIKeys = SimpleSecureStrings(
+		"key0",
+		"key1",
+	) // Use internal field for multi-key testing
 	modelCfg.Fallbacks = []string{"claude-3"}
 	models := []*ModelConfig{modelCfg}
 
@@ -196,7 +199,10 @@ func TestExpandMultiKeyModels_PreservesOtherFields(t *testing.T) {
 		RequestTimeout: 30,
 		ThinkingLevel:  "high",
 	}
-	modelCfg.APIKeys = SimpleSecureStrings("key0", "key1") // Use internal field for multi-key testing
+	modelCfg.APIKeys = SimpleSecureStrings(
+		"key0",
+		"key1",
+	) // Use internal field for multi-key testing
 	models := []*ModelConfig{modelCfg}
 
 	result := expandMultiKeyModels(models)
diff --git a/pkg/config/security.go b/pkg/config/security.go
index 79dd26e14..c31e877c2 100644
--- a/pkg/config/security.go
+++ b/pkg/config/security.go
@@ -304,11 +304,13 @@ func (s *SecureString) UnmarshalJSON(value []byte) error {
 
 func (s SecureString) MarshalYAML() (any, error) {
 	// Preserve raw value if it is already a reference (enc:// or file://)
-	if strings.HasPrefix(s.raw, credential.EncScheme) || strings.HasPrefix(s.raw, credential.FileScheme) {
+	if strings.HasPrefix(s.raw, credential.EncScheme) ||
+		strings.HasPrefix(s.raw, credential.FileScheme) {
 		return s.raw, nil
 	}
 	// If resolved is a reference format (e.g. set via Set), copy back to raw
-	if strings.HasPrefix(s.resolved, credential.EncScheme) || strings.HasPrefix(s.resolved, credential.FileScheme) {
+	if strings.HasPrefix(s.resolved, credential.EncScheme) ||
+		strings.HasPrefix(s.resolved, credential.FileScheme) {
 		s.raw = s.resolved
 		return s.raw, nil
 	}
diff --git a/pkg/config/security_integration_test.go b/pkg/config/security_integration_test.go
index 6ca8637f4..5f0b61970 100644
--- a/pkg/config/security_integration_test.go
+++ b/pkg/config/security_integration_test.go
@@ -35,7 +35,10 @@ func TestJSONUnmarshalPrivateFields(t *testing.T) {
 		t.Errorf("PublicField = %q, want 'pub'", s.PublicField)
 	}
 	if s.privateField != "" {
-		t.Errorf("privateField = %q, want empty because unexported fields are ignored", s.privateField)
+		t.Errorf(
+			"privateField = %q, want empty because unexported fields are ignored",
+			s.privateField,
+		)
 	}
 }
 
@@ -352,13 +355,21 @@ skills:
 
 		// Verify Channel tokens via Key() methods
 		// Telegram
-		assert.Equal(t, "123456789:ABCdefGHIjklMNOpqrsTUVwxyz", cfg.Channels.Telegram.Token.String())
+		assert.Equal(
+			t,
+			"123456789:ABCdefGHIjklMNOpqrsTUVwxyz",
+			cfg.Channels.Telegram.Token.String(),
+		)
 		t.Logf("Telegram Token(): %s", cfg.Channels.Telegram.Token.String())
 
 		// Feishu
 		assert.Equal(t, "feishu_test_app_secret", cfg.Channels.Feishu.AppSecret.String())
 		assert.Equal(t, "feishu_test_encrypt_key", cfg.Channels.Feishu.EncryptKey.String())
-		assert.Equal(t, "feishu_test_verification_token", cfg.Channels.Feishu.VerificationToken.String())
+		assert.Equal(
+			t,
+			"feishu_test_verification_token",
+			cfg.Channels.Feishu.VerificationToken.String(),
+		)
 		t.Logf("Feishu AppSecret(): %s", cfg.Channels.Feishu.AppSecret.String())
 		t.Logf("Feishu EncryptKey(): %s", cfg.Channels.Feishu.EncryptKey.String())
 		t.Logf("Feishu VerificationToken(): %s", cfg.Channels.Feishu.VerificationToken.String())
@@ -383,7 +394,11 @@ skills:
 
 		// LINE
 		assert.Equal(t, "line_test_channel_secret", cfg.Channels.LINE.ChannelSecret.String())
-		assert.Equal(t, "line_test_channel_access_token", cfg.Channels.LINE.ChannelAccessToken.String())
+		assert.Equal(
+			t,
+			"line_test_channel_access_token",
+			cfg.Channels.LINE.ChannelAccessToken.String(),
+		)
 		t.Logf("LINE ChannelSecret(): %s", cfg.Channels.LINE.ChannelSecret.String())
 		t.Logf("LINE ChannelAccessToken(): %s", cfg.Channels.LINE.ChannelAccessToken.String())
 
@@ -431,7 +446,11 @@ skills:
 		assert.Equal(t, "ghp-github-from-file-abc123", cfg.Tools.Skills.Github.Token.String())
 		t.Logf("Github Token(): %s", cfg.Tools.Skills.Github.Token.String())
 
-		assert.Equal(t, "clawhub-auth-token-from-file", cfg.Tools.Skills.Registries.ClawHub.AuthToken.String())
+		assert.Equal(
+			t,
+			"clawhub-auth-token-from-file",
+			cfg.Tools.Skills.Registries.ClawHub.AuthToken.String(),
+		)
 		t.Logf("ClawHub AuthToken(): %s", cfg.Tools.Skills.Registries.ClawHub.AuthToken.String())
 
 		t.Log("All security keys are successfully accessible via their respective Key() methods")
diff --git a/pkg/tools/cron.go b/pkg/tools/cron.go
index 60d9d5e5a..e205d7cf3 100644
--- a/pkg/tools/cron.go
+++ b/pkg/tools/cron.go
@@ -15,7 +15,10 @@ import (
 
 // JobExecutor is the interface for executing cron jobs through the agent
 type JobExecutor interface {
-	ProcessDirectWithChannel(ctx context.Context, content, sessionKey, channel, chatID string) (string, error)
+	ProcessDirectWithChannel(
+		ctx context.Context,
+		content, sessionKey, channel, chatID string,
+	) (string, error)
 	// PublishResponseIfNeeded sends response to the outbound bus only when the
 	// agent did not already deliver content through the message tool in this round.
 	PublishResponseIfNeeded(ctx context.Context, channel, chatID, response string)
@@ -34,8 +37,13 @@ type CronTool struct {
 // NewCronTool creates a new CronTool
 // execTimeout: 0 means no timeout, >0 sets the timeout duration
 func NewCronTool(
-	cronService *cron.CronService, executor JobExecutor, msgBus *bus.MessageBus, workspace string, restrict bool,
-	execTimeout time.Duration, config *config.Config,
+	cronService *cron.CronService,
+	executor JobExecutor,
+	msgBus *bus.MessageBus,
+	workspace string,
+	restrict bool,
+	execTimeout time.Duration,
+	config *config.Config,
 ) (*CronTool, error) {
 	allowCommand := true
 	execEnabled := true
@@ -156,7 +164,9 @@ func (t *CronTool) addJob(ctx context.Context, args map[string]any) *ToolResult
 	chatID := ToolChatID(ctx)
 
 	if channel == "" || chatID == "" {
-		return ErrorResult("no session context (channel/chat_id not set). Use this tool in an active conversation.")
+		return ErrorResult(
+			"no session context (channel/chat_id not set). Use this tool in an active conversation.",
+		)
 	}
 
 	message, ok := args["message"].(string)
@@ -208,7 +218,9 @@ func (t *CronTool) addJob(ctx context.Context, args map[string]any) *ToolResult
 	// Validate type parameter (server-side whitelist, not just LLM schema hint)
 	msgType, _ := args["type"].(string)
 	if msgType != "" && msgType != "message" && msgType != "directive" {
-		return ErrorResult(fmt.Sprintf("invalid type %q, must be 'message' or 'directive'", msgType))
+		return ErrorResult(
+			fmt.Sprintf("invalid type %q, must be 'message' or 'directive'", msgType),
+		)
 	}
 
 	// GHSA-pv8c-p6jf-3fpp: command scheduling requires internal channel. When
diff --git a/pkg/tools/cron_test.go b/pkg/tools/cron_test.go
index 186c6a75e..251db5803 100644
--- a/pkg/tools/cron_test.go
+++ b/pkg/tools/cron_test.go
@@ -49,7 +49,11 @@ func (s *stubJobExecutor) PublishResponseIfNeeded(
 	s.publishedChatID = chatID
 }
 
-func newTestCronToolWithExecutorAndConfig(t *testing.T, executor JobExecutor, cfg *config.Config) *CronTool {
+func newTestCronToolWithExecutorAndConfig(
+	t *testing.T,
+	executor JobExecutor,
+	cfg *config.Config,
+) *CronTool {
 	t.Helper()
 	storePath := filepath.Join(t.TempDir(), "cron.json")
 	cronService := cron.NewCronService(storePath, nil)
@@ -102,7 +106,10 @@ func TestCronTool_CommandDoesNotRequireConfirmByDefault(t *testing.T) {
 	})
 
 	if result.IsError {
-		t.Fatalf("expected command scheduling without confirm to succeed by default, got: %s", result.ForLLM)
+		t.Fatalf(
+			"expected command scheduling without confirm to succeed by default, got: %s",
+			result.ForLLM,
+		)
 	}
 	if !strings.Contains(result.ForLLM, "Cron job added") {
 		t.Errorf("expected 'Cron job added', got: %s", result.ForLLM)
@@ -190,7 +197,10 @@ func TestCronTool_CommandAllowedFromInternalChannel(t *testing.T) {
 	})
 
 	if result.IsError {
-		t.Fatalf("expected command scheduling to succeed from internal channel, got: %s", result.ForLLM)
+		t.Fatalf(
+			"expected command scheduling to succeed from internal channel, got: %s",
+			result.ForLLM,
+		)
 	}
 	if !strings.Contains(result.ForLLM, "Cron job added") {
 		t.Errorf("expected 'Cron job added', got: %s", result.ForLLM)
@@ -225,7 +235,10 @@ func TestCronTool_NonCommandJobAllowedFromRemoteChannel(t *testing.T) {
 	})
 
 	if result.IsError {
-		t.Fatalf("expected non-command reminder to succeed from remote channel, got: %s", result.ForLLM)
+		t.Fatalf(
+			"expected non-command reminder to succeed from remote channel, got: %s",
+			result.ForLLM,
+		)
 	}
 }
 
@@ -297,7 +310,11 @@ func TestCronTool_ExecuteJobPublishesAgentResponse(t *testing.T) {
 		t.Fatalf("sessionKey = %q, want cron-job-1", executor.lastKey)
 	}
 	if executor.lastChan != "telegram" || executor.lastChatID != "chat-1" {
-		t.Fatalf("executor target = %s/%s, want telegram/chat-1", executor.lastChan, executor.lastChatID)
+		t.Fatalf(
+			"executor target = %s/%s, want telegram/chat-1",
+			executor.lastChan,
+			executor.lastChatID,
+		)
 	}
 	if executor.lastPrompt != "send me a poem" {
 		t.Fatalf("prompt = %q, want original message", executor.lastPrompt)
@@ -306,7 +323,11 @@ func TestCronTool_ExecuteJobPublishesAgentResponse(t *testing.T) {
 		t.Fatalf("published response = %q, want generated reply", executor.publishedResp)
 	}
 	if executor.publishedChan != "telegram" || executor.publishedChatID != "chat-1" {
-		t.Fatalf("published target = %s/%s, want telegram/chat-1", executor.publishedChan, executor.publishedChatID)
+		t.Fatalf(
+			"published target = %s/%s, want telegram/chat-1",
+			executor.publishedChan,
+			executor.publishedChatID,
+		)
 	}
 }
 
@@ -342,7 +363,10 @@ func TestCronTool_ExecuteJobSkipsWhenMessageToolAlreadySent(t *testing.T) {
 	}
 
 	if executor.publishedResp != "" {
-		t.Fatalf("expected no published response when message tool already sent, got: %q", executor.publishedResp)
+		t.Fatalf(
+			"expected no published response when message tool already sent, got: %q",
+			executor.publishedResp,
+		)
 	}
 }
 
@@ -386,7 +410,9 @@ func TestCronTool_ExecuteJobDirectiveWithDeliverRoutesToAgent(t *testing.T) {
 	}
 
 	if executor.lastPrompt == "" {
-		t.Fatal("expected agent to be called for directive+deliver, but ProcessDirectWithChannel was not invoked")
+		t.Fatal(
+			"expected agent to be called for directive+deliver, but ProcessDirectWithChannel was not invoked",
+		)
 	}
 	if executor.publishedResp != "agent processed" {
 		t.Fatalf("published response = %q, want %q", executor.publishedResp, "agent processed")
diff --git a/pkg/tools/edit.go b/pkg/tools/edit.go
index d5bebf4a2..78fc512c6 100644
--- a/pkg/tools/edit.go
+++ b/pkg/tools/edit.go
@@ -16,7 +16,11 @@ type EditFileTool struct {
 }
 
 // NewEditFileTool creates a new EditFileTool with optional directory restriction.
-func NewEditFileTool(workspace string, restrict bool, allowPaths ...[]*regexp.Regexp) *EditFileTool {
+func NewEditFileTool(
+	workspace string,
+	restrict bool,
+	allowPaths ...[]*regexp.Regexp,
+) *EditFileTool {
 	var patterns []*regexp.Regexp
 	if len(allowPaths) > 0 {
 		patterns = allowPaths[0]
@@ -79,7 +83,11 @@ type AppendFileTool struct {
 	fs fileSystem
 }
 
-func NewAppendFileTool(workspace string, restrict bool, allowPaths ...[]*regexp.Regexp) *AppendFileTool {
+func NewAppendFileTool(
+	workspace string,
+	restrict bool,
+	allowPaths ...[]*regexp.Regexp,
+) *AppendFileTool {
 	var patterns []*regexp.Regexp
 	if len(allowPaths) > 0 {
 		patterns = allowPaths[0]
@@ -166,7 +174,10 @@ func replaceEditContent(content []byte, oldText, newText string) ([]byte, error)
 
 	count := strings.Count(contentStr, oldText)
 	if count > 1 {
-		return nil, fmt.Errorf("old_text appears %d times. Please provide more context to make it unique", count)
+		return nil, fmt.Errorf(
+			"old_text appears %d times. Please provide more context to make it unique",
+			count,
+		)
 	}
 
 	newContent := strings.Replace(contentStr, oldText, newText, 1)
diff --git a/pkg/tools/edit_test.go b/pkg/tools/edit_test.go
index 83a7e778c..25f89fb88 100644
--- a/pkg/tools/edit_test.go
+++ b/pkg/tools/edit_test.go
@@ -76,7 +76,8 @@ func TestEditTool_EditFile_NotFound(t *testing.T) {
 	}
 
 	// Should mention file not found
-	if !strings.Contains(result.ForLLM, "not found") && !strings.Contains(result.ForUser, "not found") {
+	if !strings.Contains(result.ForLLM, "not found") &&
+		!strings.Contains(result.ForUser, "not found") {
 		t.Errorf("Expected 'file not found' message, got ForLLM: %s", result.ForLLM)
 	}
 }
@@ -103,7 +104,8 @@ func TestEditTool_EditFile_OldTextNotFound(t *testing.T) {
 	}
 
 	// Should mention old_text not found
-	if !strings.Contains(result.ForLLM, "not found") && !strings.Contains(result.ForUser, "not found") {
+	if !strings.Contains(result.ForLLM, "not found") &&
+		!strings.Contains(result.ForUser, "not found") {
 		t.Errorf("Expected 'not found' message, got ForLLM: %s", result.ForLLM)
 	}
 }
diff --git a/pkg/tools/filesystem.go b/pkg/tools/filesystem.go
index 39d45013d..35da9ecde 100644
--- a/pkg/tools/filesystem.go
+++ b/pkg/tools/filesystem.go
@@ -20,7 +20,11 @@ import (
 
 const MaxReadFileSize = 64 * 1024 // 64KB limit to avoid context overflow
 
-func validatePathWithAllowPaths(path, workspace string, restrict bool, patterns []*regexp.Regexp) (string, error) {
+func validatePathWithAllowPaths(
+	path, workspace string,
+	restrict bool,
+	patterns []*regexp.Regexp,
+) (string, error) {
 	if workspace == "" {
 		return path, fmt.Errorf("workspace is not defined")
 	}
@@ -483,7 +487,11 @@ type WriteFileTool struct {
 	fs fileSystem
 }
 
-func NewWriteFileTool(workspace string, restrict bool, allowPaths ...[]*regexp.Regexp) *WriteFileTool {
+func NewWriteFileTool(
+	workspace string,
+	restrict bool,
+	allowPaths ...[]*regexp.Regexp,
+) *WriteFileTool {
 	var patterns []*regexp.Regexp
 	if len(allowPaths) > 0 {
 		patterns = allowPaths[0]
@@ -536,7 +544,9 @@ func (t *WriteFileTool) Execute(ctx context.Context, args map[string]any) *ToolR
 
 	if !overwrite {
 		if _, err := t.fs.Open(path); err == nil {
-			return ErrorResult(fmt.Sprintf("file: %s already exists. Set overwrite=true to replace.", path))
+			return ErrorResult(
+				fmt.Sprintf("file: %s already exists. Set overwrite=true to replace.", path),
+			)
 		}
 	}
 
diff --git a/pkg/tools/filesystem_test.go b/pkg/tools/filesystem_test.go
index 0b4dd310b..90b20b47e 100644
--- a/pkg/tools/filesystem_test.go
+++ b/pkg/tools/filesystem_test.go
@@ -59,8 +59,13 @@ func TestFilesystemTool_ReadFile_NotFound(t *testing.T) {
 	}
 
 	// Should contain error message
-	if !strings.Contains(result.ForLLM, "failed to open file") && !strings.Contains(result.ForUser, "failed to read") {
-		t.Errorf("Expected error message, got ForLLM: %s, ForUser: %s", result.ForLLM, result.ForUser)
+	if !strings.Contains(result.ForLLM, "failed to open file") &&
+		!strings.Contains(result.ForUser, "failed to read") {
+		t.Errorf(
+			"Expected error message, got ForLLM: %s, ForUser: %s",
+			result.ForLLM,
+			result.ForUser,
+		)
 	}
 }
 
@@ -78,7 +83,8 @@ func TestFilesystemTool_ReadFile_MissingPath(t *testing.T) {
 	}
 
 	// Should mention required parameter
-	if !strings.Contains(result.ForLLM, "path is required") && !strings.Contains(result.ForUser, "path is required") {
+	if !strings.Contains(result.ForLLM, "path is required") &&
+		!strings.Contains(result.ForUser, "path is required") {
 		t.Errorf("Expected 'path is required' message, got ForLLM: %s", result.ForLLM)
 	}
 }
@@ -297,7 +303,12 @@ func TestFilesystemTool_WriteFile_OverwriteSandboxed(t *testing.T) {
 		"content":   "replaced in sandbox",
 		"overwrite": true,
 	})
-	assert.False(t, result.IsError, "expected success in sandbox mode with overwrite=true, got: %s", result.ForLLM)
+	assert.False(
+		t,
+		result.IsError,
+		"expected success in sandbox mode with overwrite=true, got: %s",
+		result.ForLLM,
+	)
 
 	data, err := os.ReadFile(filepath.Join(workspace, testFile))
 	assert.NoError(t, err)
@@ -325,7 +336,8 @@ func TestFilesystemTool_ListDir_Success(t *testing.T) {
 	}
 
 	// Should list files and directories
-	if !strings.Contains(result.ForLLM, "file1.txt") || !strings.Contains(result.ForLLM, "file2.txt") {
+	if !strings.Contains(result.ForLLM, "file1.txt") ||
+		!strings.Contains(result.ForLLM, "file2.txt") {
 		t.Errorf("Expected files in listing, got: %s", result.ForLLM)
 	}
 	if !strings.Contains(result.ForLLM, "subdir") {
@@ -349,8 +361,13 @@ func TestFilesystemTool_ListDir_NotFound(t *testing.T) {
 	}
 
 	// Should contain error message
-	if !strings.Contains(result.ForLLM, "failed to read") && !strings.Contains(result.ForUser, "failed to read") {
-		t.Errorf("Expected error message, got ForLLM: %s, ForUser: %s", result.ForLLM, result.ForUser)
+	if !strings.Contains(result.ForLLM, "failed to read") &&
+		!strings.Contains(result.ForUser, "failed to read") {
+		t.Errorf(
+			"Expected error message, got ForLLM: %s, ForUser: %s",
+			result.ForLLM,
+			result.ForUser,
+		)
 	}
 }
 
@@ -397,7 +414,8 @@ func TestFilesystemTool_ReadFile_RejectsSymlinkEscape(t *testing.T) {
 	// os.Root might return different errors depending on platform/implementation
 	// but it definitely should error.
 	// Our wrapper returns "access denied or file not found"
-	if !strings.Contains(result.ForLLM, "access denied") && !strings.Contains(result.ForLLM, "file not found") &&
+	if !strings.Contains(result.ForLLM, "access denied") &&
+		!strings.Contains(result.ForLLM, "file not found") &&
 		!strings.Contains(result.ForLLM, "no such file") {
 		t.Fatalf("expected symlink escape error, got: %s", result.ForLLM)
 	}
@@ -416,10 +434,20 @@ func TestFilesystemTool_EmptyWorkspace_AccessDenied(t *testing.T) {
 	})
 
 	// We EXPECT IsError=true (access blocked due to empty workspace)
-	assert.True(t, result.IsError, "Security Regression: Empty workspace allowed access! content: %s", result.ForLLM)
+	assert.True(
+		t,
+		result.IsError,
+		"Security Regression: Empty workspace allowed access! content: %s",
+		result.ForLLM,
+	)
 
 	// Verify it failed for the right reason
-	assert.Contains(t, result.ForLLM, "workspace is not defined", "Expected 'workspace is not defined' error")
+	assert.Contains(
+		t,
+		result.ForLLM,
+		"workspace is not defined",
+		"Expected 'workspace is not defined' error",
+	)
 }
 
 // TestRootMkdirAll verifies that root.MkdirAll (used by atomicWriteFileInRoot) handles all cases:
@@ -653,7 +681,10 @@ func TestWhitelistFs_BlocksSymlinkEscapeInAllowedDir(t *testing.T) {
 	patterns := []*regexp.Regexp{regexp.MustCompile(`^` + regexp.QuoteMeta(allowedDir))}
 	tool := NewReadFileTool(workspace, true, MaxReadFileSize, patterns)
 
-	result := tool.Execute(context.Background(), map[string]any{"path": filepath.Join(linkPath, "secret.txt")})
+	result := tool.Execute(
+		context.Background(),
+		map[string]any{"path": filepath.Join(linkPath, "secret.txt")},
+	)
 	if !result.IsError {
 		t.Fatalf("expected symlink escape from allowed dir to be blocked, got: %s", result.ForLLM)
 	}
diff --git a/pkg/tools/i2c.go b/pkg/tools/i2c.go
index 779b1d5a7..e3d5c152c 100644
--- a/pkg/tools/i2c.go
+++ b/pkg/tools/i2c.go
@@ -65,7 +65,9 @@ func (t *I2CTool) Parameters() map[string]any {
 
 func (t *I2CTool) Execute(ctx context.Context, args map[string]any) *ToolResult {
 	if runtime.GOOS != "linux" {
-		return ErrorResult("I2C is only supported on Linux. This tool requires /dev/i2c-* device files.")
+		return ErrorResult(
+			"I2C is only supported on Linux. This tool requires /dev/i2c-* device files.",
+		)
 	}
 
 	action, ok := args["action"].(string)
@@ -83,7 +85,9 @@ func (t *I2CTool) Execute(ctx context.Context, args map[string]any) *ToolResult
 	case "write":
 		return t.writeDevice(args)
 	default:
-		return ErrorResult(fmt.Sprintf("unknown action: %s (valid: detect, scan, read, write)", action))
+		return ErrorResult(
+			fmt.Sprintf("unknown action: %s (valid: detect, scan, read, write)", action),
+		)
 	}
 }
 
diff --git a/pkg/tools/i2c_linux.go b/pkg/tools/i2c_linux.go
index 4eaaf8f09..ccd57b24b 100644
--- a/pkg/tools/i2c_linux.go
+++ b/pkg/tools/i2c_linux.go
@@ -55,7 +55,12 @@ func smbusProbe(fd int, addr int, hasQuick bool) bool {
 			size:      i2cSmbusQuick,
 			data:      nil,
 		}
-		_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), i2cSmbus, uintptr(unsafe.Pointer(&args)))
+		_, _, errno := syscall.Syscall(
+			syscall.SYS_IOCTL,
+			uintptr(fd),
+			i2cSmbus,
+			uintptr(unsafe.Pointer(&args)),
+		)
 		return errno == 0
 	}
 
@@ -67,7 +72,12 @@ func smbusProbe(fd int, addr int, hasQuick bool) bool {
 		size:      i2cSmbusByte,
 		data:      &data,
 	}
-	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), i2cSmbus, uintptr(unsafe.Pointer(&args)))
+	_, _, errno := syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		i2cSmbus,
+		uintptr(unsafe.Pointer(&args)),
+	)
 	return errno == 0
 }
 
@@ -83,16 +93,29 @@ func (t *I2CTool) scan(args map[string]any) *ToolResult {
 	devPath := fmt.Sprintf("/dev/i2c-%s", bus)
 	fd, err := syscall.Open(devPath, syscall.O_RDWR, 0)
 	if err != nil {
-		return ErrorResult(fmt.Sprintf("failed to open %s: %v (check permissions and i2c-dev module)", devPath, err))
+		return ErrorResult(
+			fmt.Sprintf(
+				"failed to open %s: %v (check permissions and i2c-dev module)",
+				devPath,
+				err,
+			),
+		)
 	}
 	defer syscall.Close(fd)
 
 	// Query adapter capabilities to determine available probe methods.
 	// I2C_FUNCS writes an unsigned long, which is word-sized on Linux.
 	var funcs uintptr
-	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), i2cFuncs, uintptr(unsafe.Pointer(&funcs)))
+	_, _, errno := syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		i2cFuncs,
+		uintptr(unsafe.Pointer(&funcs)),
+	)
 	if errno != 0 {
-		return ErrorResult(fmt.Sprintf("failed to query I2C adapter capabilities on %s: %v", devPath, errno))
+		return ErrorResult(
+			fmt.Sprintf("failed to query I2C adapter capabilities on %s: %v", devPath, errno),
+		)
 	}
 
 	hasQuick := funcs&i2cFuncSmbusQuick != 0
@@ -100,7 +123,10 @@ func (t *I2CTool) scan(args map[string]any) *ToolResult {
 
 	if !hasQuick && !hasReadByte {
 		return ErrorResult(
-			fmt.Sprintf("I2C adapter %s supports neither SMBus Quick nor Read Byte — cannot probe safely", devPath),
+			fmt.Sprintf(
+				"I2C adapter %s supports neither SMBus Quick nor Read Byte — cannot probe safely",
+				devPath,
+			),
 		)
 	}
 
@@ -132,7 +158,9 @@ func (t *I2CTool) scan(args map[string]any) *ToolResult {
 	}
 
 	if len(found) == 0 {
-		return SilentResult(fmt.Sprintf("No devices found on %s. Check wiring and pull-up resistors.", devPath))
+		return SilentResult(
+			fmt.Sprintf("No devices found on %s. Check wiring and pull-up resistors.", devPath),
+		)
 	}
 
 	result, _ := json.MarshalIndent(map[string]any{
diff --git a/pkg/tools/mcp_tool.go b/pkg/tools/mcp_tool.go
index 5bffb4e89..d4674d376 100644
--- a/pkg/tools/mcp_tool.go
+++ b/pkg/tools/mcp_tool.go
@@ -314,7 +314,10 @@ func (t *MCPTool) normalizeResultContent(ctx context.Context, content []mcp.Cont
 	return result
 }
 
-func (t *MCPTool) storeEmbeddedResource(ctx context.Context, content *mcp.EmbeddedResource) (string, string) {
+func (t *MCPTool) storeEmbeddedResource(
+	ctx context.Context,
+	content *mcp.EmbeddedResource,
+) (string, string) {
 	if content == nil || content.Resource == nil {
 		return "", "[MCP returned an embedded resource without data.]"
 	}
@@ -374,23 +377,39 @@ func (t *MCPTool) storeBinaryContent(
 
 	dir := media.TempDir()
 	if err := os.MkdirAll(dir, 0o700); err != nil {
-		return "", fmt.Sprintf("[MCP returned %s content (%s) but it could not be stored.]", kind, mimeType)
+		return "", fmt.Sprintf(
+			"[MCP returned %s content (%s) but it could not be stored.]",
+			kind,
+			mimeType,
+		)
 	}
 
 	ext := extensionForMIMEType(mimeType)
 	tmpFile, err := os.CreateTemp(dir, "mcp-*"+ext)
 	if err != nil {
-		return "", fmt.Sprintf("[MCP returned %s content (%s) but it could not be stored.]", kind, mimeType)
+		return "", fmt.Sprintf(
+			"[MCP returned %s content (%s) but it could not be stored.]",
+			kind,
+			mimeType,
+		)
 	}
 	tmpPath := tmpFile.Name()
 	if _, err = tmpFile.Write(data); err != nil {
 		_ = tmpFile.Close()
 		_ = os.Remove(tmpPath)
-		return "", fmt.Sprintf("[MCP returned %s content (%s) but it could not be stored.]", kind, mimeType)
+		return "", fmt.Sprintf(
+			"[MCP returned %s content (%s) but it could not be stored.]",
+			kind,
+			mimeType,
+		)
 	}
 	if err = tmpFile.Close(); err != nil {
 		_ = os.Remove(tmpPath)
-		return "", fmt.Sprintf("[MCP returned %s content (%s) but it could not be stored.]", kind, mimeType)
+		return "", fmt.Sprintf(
+			"[MCP returned %s content (%s) but it could not be stored.]",
+			kind,
+			mimeType,
+		)
 	}
 
 	scope := fmt.Sprintf(
@@ -470,7 +489,10 @@ func summarizeEmbeddedResource(content *mcp.EmbeddedResource) string {
 			normalizedMIMEType(resource.MIMEType),
 		)
 	}
-	return fmt.Sprintf("[MCP returned embedded resource (%s).]", normalizedMIMEType(resource.MIMEType))
+	return fmt.Sprintf(
+		"[MCP returned embedded resource (%s).]",
+		normalizedMIMEType(resource.MIMEType),
+	)
 }
 
 func annotationsAllowUser(annotations *mcp.Annotations) bool {
diff --git a/pkg/tools/mcp_tool_test.go b/pkg/tools/mcp_tool_test.go
index 8bbac3bc7..3b514cd82 100644
--- a/pkg/tools/mcp_tool_test.go
+++ b/pkg/tools/mcp_tool_test.go
@@ -571,7 +571,10 @@ func TestMCPTool_Execute_EmbeddedResourceBlobStoredAsMedia(t *testing.T) {
 	result := mcpTool.Execute(WithToolContext(context.Background(), "telegram", "chat-42"), nil)
 
 	if len(result.Media) != 1 {
-		t.Fatalf("expected embedded resource blob to be stored as media, got %d refs", len(result.Media))
+		t.Fatalf(
+			"expected embedded resource blob to be stored as media, got %d refs",
+			len(result.Media),
+		)
 	}
 	path, _, err := store.ResolveWithMeta(result.Media[0])
 	if err != nil {
diff --git a/pkg/tools/message_test.go b/pkg/tools/message_test.go
index 05630972e..1b8bfab4a 100644
--- a/pkg/tools/message_test.go
+++ b/pkg/tools/message_test.go
@@ -43,7 +43,10 @@ func TestMessageTool_Execute_Success(t *testing.T) {
 
 	// - ForLLM contains send status description
 	if result.ForLLM != "Message sent to test-channel:test-chat-id" {
-		t.Errorf("Expected ForLLM 'Message sent to test-channel:test-chat-id', got '%s'", result.ForLLM)
+		t.Errorf(
+			"Expected ForLLM 'Message sent to test-channel:test-chat-id', got '%s'",
+			result.ForLLM,
+		)
 	}
 
 	// - ForUser is empty (user already received message directly)
@@ -88,7 +91,10 @@ func TestMessageTool_Execute_WithCustomChannel(t *testing.T) {
 		t.Error("Expected Silent=true")
 	}
 	if result.ForLLM != "Message sent to custom-channel:custom-chat-id" {
-		t.Errorf("Expected ForLLM 'Message sent to custom-channel:custom-chat-id', got '%s'", result.ForLLM)
+		t.Errorf(
+			"Expected ForLLM 'Message sent to custom-channel:custom-chat-id', got '%s'",
+			result.ForLLM,
+		)
 	}
 }
 
diff --git a/pkg/tools/normalization.go b/pkg/tools/normalization.go
index 3a76c5d92..9cd9c65c6 100644
--- a/pkg/tools/normalization.go
+++ b/pkg/tools/normalization.go
@@ -215,28 +215,43 @@ func storeInlineDataURL(
 	payload = strings.NewReplacer("\n", "", "\r", "", "\t", "", " ", "").Replace(payload)
 	decoded, err := base64.StdEncoding.DecodeString(payload)
 	if err != nil {
-		return "", fmt.Sprintf("[Tool returned inline media content (%s) that could not be decoded.]", mimeType)
+		return "", fmt.Sprintf(
+			"[Tool returned inline media content (%s) that could not be decoded.]",
+			mimeType,
+		)
 	}
 
 	dir := media.TempDir()
 	if err = os.MkdirAll(dir, 0o700); err != nil {
-		return "", fmt.Sprintf("[Tool returned inline media content (%s) but it could not be stored.]", mimeType)
+		return "", fmt.Sprintf(
+			"[Tool returned inline media content (%s) but it could not be stored.]",
+			mimeType,
+		)
 	}
 
 	ext := extensionForMIMEType(mimeType)
 	tmpFile, err := os.CreateTemp(dir, "tool-inline-*"+ext)
 	if err != nil {
-		return "", fmt.Sprintf("[Tool returned inline media content (%s) but it could not be stored.]", mimeType)
+		return "", fmt.Sprintf(
+			"[Tool returned inline media content (%s) but it could not be stored.]",
+			mimeType,
+		)
 	}
 	tmpPath := tmpFile.Name()
 	if _, err = tmpFile.Write(decoded); err != nil {
 		tmpFile.Close()
 		_ = os.Remove(tmpPath)
-		return "", fmt.Sprintf("[Tool returned inline media content (%s) but it could not be stored.]", mimeType)
+		return "", fmt.Sprintf(
+			"[Tool returned inline media content (%s) but it could not be stored.]",
+			mimeType,
+		)
 	}
 	if err = tmpFile.Close(); err != nil {
 		_ = os.Remove(tmpPath)
-		return "", fmt.Sprintf("[Tool returned inline media content (%s) but it could not be stored.]", mimeType)
+		return "", fmt.Sprintf(
+			"[Tool returned inline media content (%s) but it could not be stored.]",
+			mimeType,
+		)
 	}
 
 	filename := sanitizeIdentifierComponent(toolName) + ext
@@ -255,7 +270,10 @@ func storeInlineDataURL(
 	}, scope)
 	if err != nil {
 		_ = os.Remove(tmpPath)
-		return "", fmt.Sprintf("[Tool returned inline media content (%s) but it could not be registered.]", mimeType)
+		return "", fmt.Sprintf(
+			"[Tool returned inline media content (%s) but it could not be registered.]",
+			mimeType,
+		)
 	}
 
 	return ref, fmt.Sprintf(inlineMediaStoredMessage, mimeType)
diff --git a/pkg/tools/registry.go b/pkg/tools/registry.go
index 56af8d695..e16be0ccb 100644
--- a/pkg/tools/registry.go
+++ b/pkg/tools/registry.go
@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"sort"
+	"strings"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -24,6 +25,7 @@ type ToolRegistry struct {
 	mu         sync.RWMutex
 	version    atomic.Uint64 // incremented on Register/RegisterHidden for cache invalidation
 	mediaStore media.MediaStore
+	allowlist  map[string]struct{}
 }
 
 type mediaStoreAware interface {
@@ -36,10 +38,40 @@ func NewToolRegistry() *ToolRegistry {
 	}
 }
 
+// SetAllowlist restricts registrations to the provided runtime tool names.
+// A nil slice means "allow all". An empty-but-non-nil slice means "allow none".
+func (r *ToolRegistry) SetAllowlist(names []string) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	if names == nil {
+		r.allowlist = nil
+		return
+	}
+
+	allowlist := make(map[string]struct{}, len(names))
+	for _, name := range names {
+		trimmed := strings.TrimSpace(name)
+		if trimmed == "" {
+			continue
+		}
+		allowlist[trimmed] = struct{}{}
+	}
+	r.allowlist = allowlist
+}
+
 func (r *ToolRegistry) Register(tool Tool) {
 	r.mu.Lock()
 	defer r.mu.Unlock()
 	name := tool.Name()
+	if !r.toolAllowedLocked(name) {
+		logger.DebugCF(
+			"tools",
+			"Skipped core tool registration by agent allowlist",
+			map[string]any{"name": name},
+		)
+		return
+	}
 	if _, exists := r.tools[name]; exists {
 		logger.WarnCF("tools", "Tool registration overwrites existing tool",
 			map[string]any{"name": name})
@@ -61,6 +93,14 @@ func (r *ToolRegistry) RegisterHidden(tool Tool) {
 	r.mu.Lock()
 	defer r.mu.Unlock()
 	name := tool.Name()
+	if !r.toolAllowedLocked(name) {
+		logger.DebugCF(
+			"tools",
+			"Skipped hidden tool registration by agent allowlist",
+			map[string]any{"name": name},
+		)
+		return
+	}
 	if _, exists := r.tools[name]; exists {
 		logger.WarnCF("tools", "Hidden tool registration overwrites existing tool",
 			map[string]any{"name": name})
@@ -128,6 +168,14 @@ func (r *ToolRegistry) Version() uint64 {
 	return r.version.Load()
 }
 
+func (r *ToolRegistry) toolAllowedLocked(name string) bool {
+	if r.allowlist == nil {
+		return true
+	}
+	_, ok := r.allowlist[name]
+	return ok
+}
+
 // HiddenToolSnapshot holds a consistent snapshot of hidden tools and the
 // registry version at which it was taken. Used by BM25SearchTool cache.
 type HiddenToolSnapshot struct {
@@ -203,7 +251,9 @@ func (r *ToolRegistry) ExecuteWithContext(
 			map[string]any{
 				"tool": name,
 			})
-		return ErrorResult(fmt.Sprintf("tool %q not found", name)).WithError(fmt.Errorf("tool not found"))
+		return ErrorResult(
+			fmt.Sprintf("tool %q not found", name),
+		).WithError(fmt.Errorf("tool not found"))
 	}
 
 	// Validate arguments against the tool's declared schema.
@@ -385,6 +435,12 @@ func (r *ToolRegistry) Clone() *ToolRegistry {
 		tools:      make(map[string]*ToolEntry, len(r.tools)),
 		mediaStore: r.mediaStore,
 	}
+	if r.allowlist != nil {
+		clone.allowlist = make(map[string]struct{}, len(r.allowlist))
+		for name := range r.allowlist {
+			clone.allowlist[name] = struct{}{}
+		}
+	}
 	for name, entry := range r.tools {
 		clone.tools[name] = &ToolEntry{
 			Tool:   entry.Tool,
@@ -417,7 +473,10 @@ func (r *ToolRegistry) GetSummaries() []string {
 			continue
 		}
 
-		summaries = append(summaries, fmt.Sprintf("- `%s` - %s", entry.Tool.Name(), entry.Tool.Description()))
+		summaries = append(
+			summaries,
+			fmt.Sprintf("- `%s` - %s", entry.Tool.Name(), entry.Tool.Description()),
+		)
 	}
 	return summaries
 }
diff --git a/pkg/tools/registry_test.go b/pkg/tools/registry_test.go
index db52749f6..17b3cd127 100644
--- a/pkg/tools/registry_test.go
+++ b/pkg/tools/registry_test.go
@@ -44,7 +44,11 @@ type mockAsyncRegistryTool struct {
 	lastCB AsyncCallback
 }
 
-func (m *mockAsyncRegistryTool) ExecuteAsync(_ context.Context, args map[string]any, cb AsyncCallback) *ToolResult {
+func (m *mockAsyncRegistryTool) ExecuteAsync(
+	_ context.Context,
+	args map[string]any,
+	cb AsyncCallback,
+) *ToolResult {
 	m.lastCB = cb
 	return m.result
 }
@@ -95,6 +99,28 @@ func TestToolRegistry_RegisterAndGet(t *testing.T) {
 	}
 }
 
+func TestToolRegistry_AllowlistFiltersRegistrations(t *testing.T) {
+	r := NewToolRegistry()
+	r.SetAllowlist([]string{"allowed_tool"})
+
+	r.Register(newMockTool("allowed_tool", "allowed"))
+	r.Register(newMockTool("blocked_tool", "blocked"))
+	r.RegisterHidden(newMockTool("hidden_blocked", "hidden blocked"))
+
+	if _, ok := r.Get("allowed_tool"); !ok {
+		t.Fatal("expected allowed_tool to be registered")
+	}
+	if _, ok := r.Get("blocked_tool"); ok {
+		t.Fatal("blocked_tool should not be registered")
+	}
+	if _, ok := r.Get("hidden_blocked"); ok {
+		t.Fatal("hidden_blocked should not be registered")
+	}
+	if got := r.List(); len(got) != 1 || got[0] != "allowed_tool" {
+		t.Fatalf("registry list = %v, want [allowed_tool]", got)
+	}
+}
+
 func TestToolRegistry_Get_NotFound(t *testing.T) {
 	r := NewToolRegistry()
 	_, ok := r.Get("nonexistent")
@@ -269,7 +295,11 @@ func TestToolRegistry_ToProviderDefs(t *testing.T) {
 		t.Errorf("Name: want %q, got %q", want.Function.Name, got.Function.Name)
 	}
 	if got.Function.Description != want.Function.Description {
-		t.Errorf("Description: want %q, got %q", want.Function.Description, got.Function.Description)
+		t.Errorf(
+			"Description: want %q, got %q",
+			want.Function.Description,
+			got.Function.Description,
+		)
 	}
 }
 
@@ -372,7 +402,10 @@ func TestToolRegistry_Clone(t *testing.T) {
 		t.Errorf("expected parent to have 4 tools, got %d", r.Count())
 	}
 	if clone.Count() != 3 {
-		t.Errorf("expected clone to still have 3 tools after parent mutation, got %d", clone.Count())
+		t.Errorf(
+			"expected clone to still have 3 tools after parent mutation, got %d",
+			clone.Count(),
+		)
 	}
 	if _, ok := clone.Get("spawn"); ok {
 		t.Error("expected clone NOT to have 'spawn' tool registered on parent after cloning")
@@ -668,7 +701,14 @@ func TestToolRegistry_ExecuteWithContext_SanitizesLargeBase64Payload(t *testing.
 		result: SilentResult(payload),
 	})
 
-	result := r.ExecuteWithContext(context.Background(), "base64_tool", nil, "telegram", "chat-1", nil)
+	result := r.ExecuteWithContext(
+		context.Background(),
+		"base64_tool",
+		nil,
+		"telegram",
+		"chat-1",
+		nil,
+	)
 
 	if result.ForLLM != largeBase64OmittedMessage {
 		t.Fatalf("expected sanitized payload, got %q", result.ForLLM)
@@ -688,7 +728,14 @@ func TestToolRegistry_ExecuteWithContext_ExtractsInlineMediaDataURL(t *testing.T
 		result: SilentResult(payload),
 	})
 
-	result := r.ExecuteWithContext(context.Background(), "inline_media_tool", nil, "telegram", "chat-42", nil)
+	result := r.ExecuteWithContext(
+		context.Background(),
+		"inline_media_tool",
+		nil,
+		"telegram",
+		"chat-42",
+		nil,
+	)
 
 	if len(result.Media) != 1 {
 		t.Fatalf("expected 1 media ref, got %d", len(result.Media))
@@ -723,7 +770,14 @@ func TestToolRegistry_ExecuteWithContext_SanitizesInlineMediaWithoutStore(t *tes
 		result: SilentResult(payload),
 	})
 
-	result := r.ExecuteWithContext(context.Background(), "inline_media_no_store", nil, "telegram", "chat-42", nil)
+	result := r.ExecuteWithContext(
+		context.Background(),
+		"inline_media_no_store",
+		nil,
+		"telegram",
+		"chat-42",
+		nil,
+	)
 
 	if strings.Contains(result.ForLLM, "data:image/png;base64") {
 		t.Fatalf("expected inline data URL to be removed from ForLLM, got %q", result.ForLLM)
diff --git a/pkg/tools/result.go b/pkg/tools/result.go
index c81213125..1976eb88b 100644
--- a/pkg/tools/result.go
+++ b/pkg/tools/result.go
@@ -80,7 +80,10 @@ func (tr *ToolResult) ContentForLLM() string {
 		}
 	}
 	if len(tr.ArtifactTags) > 0 {
-		artifactNote := "Local artifact paths: " + strings.Join(tr.ArtifactTags, " ") + "\n" + artifactPathsLLMNote
+		artifactNote := "Local artifact paths: " + strings.Join(
+			tr.ArtifactTags,
+			" ",
+		) + "\n" + artifactPathsLLMNote
 		if content == "" {
 			content = artifactNote
 		} else if !strings.Contains(content, artifactNote) {
diff --git a/pkg/tools/result_test.go b/pkg/tools/result_test.go
index 5f08cb4fa..87b2f1b4b 100644
--- a/pkg/tools/result_test.go
+++ b/pkg/tools/result_test.go
@@ -142,7 +142,11 @@ func TestToolResultJSONSerialization(t *testing.T) {
 				t.Errorf("ForLLM mismatch: got '%s', want '%s'", decoded.ForLLM, tt.result.ForLLM)
 			}
 			if decoded.ForUser != tt.result.ForUser {
-				t.Errorf("ForUser mismatch: got '%s', want '%s'", decoded.ForUser, tt.result.ForUser)
+				t.Errorf(
+					"ForUser mismatch: got '%s', want '%s'",
+					decoded.ForUser,
+					tt.result.ForUser,
+				)
 			}
 			if decoded.Silent != tt.result.Silent {
 				t.Errorf("Silent mismatch: got %v, want %v", decoded.Silent, tt.result.Silent)
diff --git a/pkg/tools/search_tool.go b/pkg/tools/search_tool.go
index f41c80d90..21326504d 100644
--- a/pkg/tools/search_tool.go
+++ b/pkg/tools/search_tool.go
@@ -56,19 +56,38 @@ func (t *RegexSearchTool) Execute(ctx context.Context, args map[string]any) *Too
 	}
 
 	if len(pattern) > MaxRegexPatternLength {
-		logger.WarnCF("discovery", "Regex pattern rejected (too long)", map[string]any{"len": len(pattern)})
-		return ErrorResult(fmt.Sprintf("Pattern too long: max %d characters allowed", MaxRegexPatternLength))
+		logger.WarnCF(
+			"discovery",
+			"Regex pattern rejected (too long)",
+			map[string]any{"len": len(pattern)},
+		)
+		return ErrorResult(
+			fmt.Sprintf("Pattern too long: max %d characters allowed", MaxRegexPatternLength),
+		)
 	}
 
 	logger.DebugCF("discovery", "Regex search", map[string]any{"pattern": pattern})
 
 	res, err := t.registry.SearchRegex(pattern, t.maxSearchResults)
 	if err != nil {
-		logger.WarnCF("discovery", "Invalid regex pattern", map[string]any{"pattern": pattern, "error": err.Error()})
-		return ErrorResult(fmt.Sprintf("Invalid regex pattern syntax: %v. Please fix your regex and try again.", err))
+		logger.WarnCF(
+			"discovery",
+			"Invalid regex pattern",
+			map[string]any{"pattern": pattern, "error": err.Error()},
+		)
+		return ErrorResult(
+			fmt.Sprintf(
+				"Invalid regex pattern syntax: %v. Please fix your regex and try again.",
+				err,
+			),
+		)
 	}
 
-	logger.InfoCF("discovery", "Regex search completed", map[string]any{"pattern": pattern, "results": len(res)})
+	logger.InfoCF(
+		"discovery",
+		"Regex search completed",
+		map[string]any{"pattern": pattern, "results": len(res)},
+	)
 	return formatDiscoveryResponse(t.registry, res, t.ttl)
 }
 
@@ -138,7 +157,11 @@ func (t *BM25SearchTool) Execute(ctx context.Context, args map[string]any) *Tool
 		}
 	}
 
-	logger.InfoCF("discovery", "BM25 search completed", map[string]any{"query": query, "results": len(results)})
+	logger.InfoCF(
+		"discovery",
+		"BM25 search completed",
+		map[string]any{"query": query, "results": len(results)},
+	)
 	return formatDiscoveryResponse(t.registry, results, t.ttl)
 }
 
@@ -150,7 +173,10 @@ type ToolSearchResult struct {
 	Description string `json:"description"`
 }
 
-func (r *ToolRegistry) SearchRegex(pattern string, maxSearchResults int) ([]ToolSearchResult, error) {
+func (r *ToolRegistry) SearchRegex(
+	pattern string,
+	maxSearchResults int,
+) ([]ToolSearchResult, error) {
 	if maxSearchResults <= 0 {
 		return nil, nil
 	}
@@ -188,7 +214,11 @@ func (r *ToolRegistry) SearchRegex(pattern string, maxSearchResults int) ([]Tool
 	return results, nil
 }
 
-func formatDiscoveryResponse(registry *ToolRegistry, results []ToolSearchResult, ttl int) *ToolResult {
+func formatDiscoveryResponse(
+	registry *ToolRegistry,
+	results []ToolSearchResult,
+	ttl int,
+) *ToolResult {
 	if len(results) == 0 {
 		return SilentResult("No tools found matching the query.")
 	}
@@ -274,7 +304,11 @@ func (t *BM25SearchTool) getOrBuildEngine() *bm25CachedEngine {
 	cached := &bm25CachedEngine{engine: buildBM25Engine(docs)}
 	t.cachedEngine = cached
 	t.cacheVersion = snap.Version
-	logger.DebugCF("discovery", "BM25 engine rebuilt", map[string]any{"docs": len(docs), "version": snap.Version})
+	logger.DebugCF(
+		"discovery",
+		"BM25 engine rebuilt",
+		map[string]any{"docs": len(docs), "version": snap.Version},
+	)
 	return cached
 }
 
diff --git a/pkg/tools/search_tools_test.go b/pkg/tools/search_tools_test.go
index 3aae941cb..72cb11444 100644
--- a/pkg/tools/search_tools_test.go
+++ b/pkg/tools/search_tools_test.go
@@ -93,7 +93,10 @@ func TestRegexSearchTool_Execute(t *testing.T) {
 		reg.mu.RLock()
 		defer reg.mu.RUnlock()
 		if reg.tools["mcp_read_file"].TTL != 5 {
-			t.Errorf("Expected TTL of 'mcp_read_file' to be promoted to 5, got %d", reg.tools["mcp_read_file"].TTL)
+			t.Errorf(
+				"Expected TTL of 'mcp_read_file' to be promoted to 5, got %d",
+				reg.tools["mcp_read_file"].TTL,
+			)
 		}
 		if reg.tools["mcp_fetch_net"].TTL != 0 {
 			t.Errorf("Expected 'mcp_fetch_net' to NOT be promoted (TTL=0)")
diff --git a/pkg/tools/send_file.go b/pkg/tools/send_file.go
index 44198381e..a344f4b5c 100644
--- a/pkg/tools/send_file.go
+++ b/pkg/tools/send_file.go
@@ -142,7 +142,10 @@ func (t *SendFileTool) Execute(ctx context.Context, args map[string]any) *ToolRe
 		return ErrorResult(fmt.Sprintf("failed to register media: %v", err))
 	}
 
-	return MediaResult(fmt.Sprintf("File %q sent to user", filename), []string{ref}).WithResponseHandled()
+	return MediaResult(
+		fmt.Sprintf("File %q sent to user", filename),
+		[]string{ref},
+	).WithResponseHandled()
 }
 
 // detectMediaType determines the MIME type of a file.
diff --git a/pkg/tools/send_file_test.go b/pkg/tools/send_file_test.go
index f36baf7d0..26b3c17ab 100644
--- a/pkg/tools/send_file_test.go
+++ b/pkg/tools/send_file_test.go
@@ -79,7 +79,11 @@ func TestSendFileTool_FileTooLarge(t *testing.T) {
 func TestSendFileTool_DefaultMaxSize(t *testing.T) {
 	tool := NewSendFileTool("/tmp", false, 0, nil)
 	if tool.maxFileSize != config.DefaultMaxMediaSize {
-		t.Errorf("expected default max size %d, got %d", config.DefaultMaxMediaSize, tool.maxFileSize)
+		t.Errorf(
+			"expected default max size %d, got %d",
+			config.DefaultMaxMediaSize,
+			tool.maxFileSize,
+		)
 	}
 }
 
@@ -162,7 +166,11 @@ func TestSendFileTool_AllowsWhitelistedMediaTempPath(t *testing.T) {
 	t.Cleanup(func() { _ = os.Remove(testPath) })
 
 	pattern := regexp.MustCompile(
-		"^" + regexp.QuoteMeta(filepath.Clean(mediaDir)) + "(?:" + regexp.QuoteMeta(string(os.PathSeparator)) + "|$)",
+		"^" + regexp.QuoteMeta(
+			filepath.Clean(mediaDir),
+		) + "(?:" + regexp.QuoteMeta(
+			string(os.PathSeparator),
+		) + "|$)",
 	)
 
 	store := media.NewFileMediaStore()
diff --git a/pkg/tools/shell.go b/pkg/tools/shell.go
index 6ee1cb993..0d1c4c5db 100644
--- a/pkg/tools/shell.go
+++ b/pkg/tools/shell.go
@@ -113,7 +113,11 @@ var (
 	}
 )
 
-func NewExecTool(workingDir string, restrict bool, allowPaths ...[]*regexp.Regexp) (*ExecTool, error) {
+func NewExecTool(
+	workingDir string,
+	restrict bool,
+	allowPaths ...[]*regexp.Regexp,
+) (*ExecTool, error) {
 	return NewExecToolWithConfig(workingDir, restrict, nil, allowPaths...)
 }
 
@@ -193,8 +197,16 @@ func (t *ExecTool) Parameters() map[string]any {
 		"type": "object",
 		"properties": map[string]any{
 			"action": map[string]any{
-				"type":        "string",
-				"enum":        []string{"run", "list", "poll", "read", "write", "kill", "send-keys"},
+				"type": "string",
+				"enum": []string{
+					"run",
+					"list",
+					"poll",
+					"read",
+					"write",
+					"kill",
+					"send-keys",
+				},
 				"description": "Action: run (execute command), list (show sessions), poll (check status), read (get output), write (send input), kill (terminate), send-keys (send keys to PTY)",
 			},
 			"command": map[string]any{
@@ -300,7 +312,12 @@ func (t *ExecTool) executeRun(ctx context.Context, args map[string]any) *ToolRes
 	cwd := t.workingDir
 	if wd, ok := args["cwd"].(string); ok && wd != "" {
 		if t.restrictToWorkspace && t.workingDir != "" {
-			resolvedWD, err := validatePathWithAllowPaths(wd, t.workingDir, true, t.allowedPathPatterns)
+			resolvedWD, err := validatePathWithAllowPaths(
+				wd,
+				t.workingDir,
+				true,
+				t.allowedPathPatterns,
+			)
 			if err != nil {
 				return ErrorResult("Command blocked by safety guard (" + err.Error() + ")")
 			}
@@ -326,7 +343,9 @@ func (t *ExecTool) executeRun(ctx context.Context, args map[string]any) *ToolRes
 	if t.restrictToWorkspace && t.workingDir != "" && cwd != t.workingDir {
 		resolved, err := filepath.EvalSymlinks(cwd)
 		if err != nil {
-			return ErrorResult(fmt.Sprintf("Command blocked by safety guard (path resolution failed: %v)", err))
+			return ErrorResult(
+				fmt.Sprintf("Command blocked by safety guard (path resolution failed: %v)", err),
+			)
 		}
 		if isAllowedPath(resolved, t.allowedPathPatterns) {
 			cwd = resolved
@@ -364,7 +383,14 @@ func (t *ExecTool) runSync(ctx context.Context, command, cwd string) *ToolResult
 
 	var cmd *exec.Cmd
 	if runtime.GOOS == "windows" {
-		cmd = exec.CommandContext(cmdCtx, "powershell", "-NoProfile", "-NonInteractive", "-Command", command)
+		cmd = exec.CommandContext(
+			cmdCtx,
+			"powershell",
+			"-NoProfile",
+			"-NonInteractive",
+			"-Command",
+			command,
+		)
 	} else {
 		cmd = exec.CommandContext(cmdCtx, "sh", "-c", command)
 	}
@@ -442,7 +468,10 @@ func (t *ExecTool) runSync(ctx context.Context, command, cwd string) *ToolResult
 
 	maxLen := 10000
 	if len(output) > maxLen {
-		output = output[:maxLen] + fmt.Sprintf("\n... (truncated, %d more chars)", len(output)-maxLen)
+		output = output[:maxLen] + fmt.Sprintf(
+			"\n... (truncated, %d more chars)",
+			len(output)-maxLen,
+		)
 	}
 
 	if err != nil {
@@ -460,7 +489,11 @@ func (t *ExecTool) runSync(ctx context.Context, command, cwd string) *ToolResult
 	}
 }
 
-func (t *ExecTool) runBackground(ctx context.Context, command, cwd string, ptyEnabled bool) *ToolResult {
+func (t *ExecTool) runBackground(
+	ctx context.Context,
+	command, cwd string,
+	ptyEnabled bool,
+) *ToolResult {
 	sessionID := generateSessionID()
 	session := &ProcessSession{
 		ID:         sessionID,
@@ -553,7 +586,8 @@ func (t *ExecTool) runBackground(ctx context.Context, command, cwd string, ptyEn
 				n, err := session.ptyMaster.Read(buf)
 				if n > 0 {
 					raw := string(buf[:n])
-					if mode := detectPtyKeyMode(raw); mode != PtyKeyModeNotFound && mode != session.GetPtyKeyMode() {
+					if mode := detectPtyKeyMode(raw); mode != PtyKeyModeNotFound &&
+						mode != session.GetPtyKeyMode() {
 						session.SetPtyKeyMode(mode)
 					}
 
@@ -734,12 +768,16 @@ func (t *ExecTool) executeWrite(args map[string]any) *ToolResult {
 	}
 
 	if session.IsDone() {
-		return ErrorResult(fmt.Sprintf("process already exited with code %d", session.GetExitCode()))
+		return ErrorResult(
+			fmt.Sprintf("process already exited with code %d", session.GetExitCode()),
+		)
 	}
 
 	if err := session.Write(data); err != nil {
 		if errors.Is(err, ErrSessionDone) {
-			return ErrorResult(fmt.Sprintf("process already exited with code %d", session.GetExitCode()))
+			return ErrorResult(
+				fmt.Sprintf("process already exited with code %d", session.GetExitCode()),
+			)
 		}
 		return ErrorResult(fmt.Sprintf("failed to write to session: %v", err))
 	}
@@ -770,7 +808,9 @@ func (t *ExecTool) executeKill(args map[string]any) *ToolResult {
 	}
 
 	if session.IsDone() {
-		return ErrorResult(fmt.Sprintf("process already exited with code %d", session.GetExitCode()))
+		return ErrorResult(
+			fmt.Sprintf("process already exited with code %d", session.GetExitCode()),
+		)
 	}
 
 	if err := session.Kill(); err != nil {
@@ -992,12 +1032,16 @@ func (t *ExecTool) executeSendKeys(args map[string]any) *ToolResult {
 	}
 
 	if session.IsDone() {
-		return ErrorResult(fmt.Sprintf("process already exited with code %d", session.GetExitCode()))
+		return ErrorResult(
+			fmt.Sprintf("process already exited with code %d", session.GetExitCode()),
+		)
 	}
 
 	if err := session.Write(data); err != nil {
 		if errors.Is(err, ErrSessionDone) {
-			return ErrorResult(fmt.Sprintf("process already exited with code %d", session.GetExitCode()))
+			return ErrorResult(
+				fmt.Sprintf("process already exited with code %d", session.GetExitCode()),
+			)
 		}
 		return ErrorResult(fmt.Sprintf("failed to send keys: %v", err))
 	}
diff --git a/pkg/tools/shell_test.go b/pkg/tools/shell_test.go
index a8de2f4c9..228ec1067 100644
--- a/pkg/tools/shell_test.go
+++ b/pkg/tools/shell_test.go
@@ -100,8 +100,13 @@ func TestShellTool_Timeout(t *testing.T) {
 	}
 
 	// Should mention timeout
-	if !strings.Contains(result.ForLLM, "timed out") && !strings.Contains(result.ForUser, "timed out") {
-		t.Errorf("Expected timeout message, got ForLLM: %s, ForUser: %s", result.ForLLM, result.ForUser)
+	if !strings.Contains(result.ForLLM, "timed out") &&
+		!strings.Contains(result.ForUser, "timed out") {
+		t.Errorf(
+			"Expected timeout message, got ForLLM: %s, ForUser: %s",
+			result.ForLLM,
+			result.ForUser,
+		)
 	}
 }
 
@@ -156,7 +161,11 @@ func TestShellTool_DangerousCommand(t *testing.T) {
 	}
 
 	if !strings.Contains(result.ForLLM, "blocked") && !strings.Contains(result.ForUser, "blocked") {
-		t.Errorf("Expected 'blocked' message, got ForLLM: %s, ForUser: %s", result.ForLLM, result.ForUser)
+		t.Errorf(
+			"Expected 'blocked' message, got ForLLM: %s, ForUser: %s",
+			result.ForLLM,
+			result.ForUser,
+		)
 	}
 }
 
@@ -177,7 +186,11 @@ func TestShellTool_DangerousCommand_KillBlocked(t *testing.T) {
 		t.Errorf("Expected kill command to be blocked")
 	}
 	if !strings.Contains(result.ForLLM, "blocked") && !strings.Contains(result.ForUser, "blocked") {
-		t.Errorf("Expected blocked message, got ForLLM: %s, ForUser: %s", result.ForLLM, result.ForUser)
+		t.Errorf(
+			"Expected blocked message, got ForLLM: %s, ForUser: %s",
+			result.ForLLM,
+			result.ForUser,
+		)
 	}
 }
 
@@ -269,7 +282,10 @@ func TestShellTool_WorkingDir_OutsideWorkspace(t *testing.T) {
 	})
 
 	if !result.IsError {
-		t.Fatalf("expected working_dir outside workspace to be blocked, got output: %s", result.ForLLM)
+		t.Fatalf(
+			"expected working_dir outside workspace to be blocked, got output: %s",
+			result.ForLLM,
+		)
 	}
 	if !strings.Contains(result.ForLLM, "blocked") {
 		t.Errorf("expected 'blocked' in error, got: %s", result.ForLLM)
@@ -444,7 +460,10 @@ func TestShellTool_DevNullAllowed(t *testing.T) {
 	}
 
 	for _, cmd := range commands {
-		result := tool.Execute(context.Background(), map[string]any{"action": "run", "command": cmd})
+		result := tool.Execute(
+			context.Background(),
+			map[string]any{"action": "run", "command": cmd},
+		)
 		if result.IsError && strings.Contains(result.ForLLM, "blocked") {
 			t.Errorf("command should not be blocked: %s\n  error: %s", cmd, result.ForLLM)
 		}
@@ -473,7 +492,10 @@ func TestShellTool_BlockDevices(t *testing.T) {
 	}
 
 	for _, cmd := range blocked {
-		result := tool.Execute(context.Background(), map[string]any{"action": "run", "command": cmd})
+		result := tool.Execute(
+			context.Background(),
+			map[string]any{"action": "run", "command": cmd},
+		)
 		if !result.IsError {
 			t.Errorf("expected block device write to be blocked: %s", cmd)
 		}
@@ -497,9 +519,16 @@ func TestShellTool_SafePathsInWorkspaceRestriction(t *testing.T) {
 	}
 
 	for _, cmd := range commands {
-		result := tool.Execute(context.Background(), map[string]any{"action": "run", "command": cmd})
+		result := tool.Execute(
+			context.Background(),
+			map[string]any{"action": "run", "command": cmd},
+		)
 		if result.IsError && strings.Contains(result.ForLLM, "path outside working dir") {
-			t.Errorf("safe path should not be blocked by workspace check: %s\n  error: %s", cmd, result.ForLLM)
+			t.Errorf(
+				"safe path should not be blocked by workspace check: %s\n  error: %s",
+				cmd,
+				result.ForLLM,
+			)
 		}
 	}
 }
@@ -591,7 +620,10 @@ func TestShellTool_CustomAllowPatterns(t *testing.T) {
 		"command": "git push origin main",
 	})
 	if result.IsError && strings.Contains(result.ForLLM, "blocked") {
-		t.Errorf("custom allow pattern should exempt 'git push origin main', got: %s", result.ForLLM)
+		t.Errorf(
+			"custom allow pattern should exempt 'git push origin main', got: %s",
+			result.ForLLM,
+		)
 	}
 
 	// "git push upstream main" should still be blocked (does not match allow pattern).
@@ -629,7 +661,11 @@ func TestShellTool_URLsNotBlocked(t *testing.T) {
 		result := tool.Execute(ctx, map[string]any{"action": "run", "command": cmd})
 		cancel()
 		if result.IsError && strings.Contains(result.ForLLM, "path outside working dir") {
-			t.Errorf("command with URL should not be blocked by workspace check: %s\n  error: %s", cmd, result.ForLLM)
+			t.Errorf(
+				"command with URL should not be blocked by workspace check: %s\n  error: %s",
+				cmd,
+				result.ForLLM,
+			)
 		}
 	}
 }
@@ -652,7 +688,10 @@ func TestShellTool_FileURISandboxing(t *testing.T) {
 	}
 
 	for _, cmd := range blockedCommands {
-		result := tool.Execute(context.Background(), map[string]any{"action": "run", "command": cmd})
+		result := tool.Execute(
+			context.Background(),
+			map[string]any{"action": "run", "command": cmd},
+		)
 		if !result.IsError || !strings.Contains(result.ForLLM, "path outside working dir") {
 			t.Errorf("file:// URI outside workspace should be blocked: %s", cmd)
 		}
@@ -670,9 +709,16 @@ func TestShellTool_FileURISandboxing(t *testing.T) {
 	}
 
 	for _, cmd := range allowedCommands {
-		result := tool.Execute(context.Background(), map[string]any{"action": "run", "command": cmd})
+		result := tool.Execute(
+			context.Background(),
+			map[string]any{"action": "run", "command": cmd},
+		)
 		if result.IsError && strings.Contains(result.ForLLM, "path outside working dir") {
-			t.Errorf("file:// URI inside workspace should be allowed: %s\n  error: %s", cmd, result.ForLLM)
+			t.Errorf(
+				"file:// URI inside workspace should be allowed: %s\n  error: %s",
+				cmd,
+				result.ForLLM,
+			)
 		}
 	}
 }
@@ -696,7 +742,10 @@ func TestShellTool_URLBypassPrevented(t *testing.T) {
 	}
 
 	for _, cmd := range blockedCommands {
-		result := tool.Execute(context.Background(), map[string]any{"action": "run", "command": cmd})
+		result := tool.Execute(
+			context.Background(),
+			map[string]any{"action": "run", "command": cmd},
+		)
 		if !result.IsError || !strings.Contains(result.ForLLM, "path outside working dir") {
 			t.Errorf("bypass attempt should be blocked: %q\n  got: %s", cmd, result.ForLLM)
 		}
@@ -1221,7 +1270,9 @@ func TestShellTool_PTY_ProcessGroupKill(t *testing.T) {
 	// The binary is created in /tmp/test_pgroup.c and compiled as part of test setup.
 	testBinary := "/tmp/test_pgroup"
 	if _, err := os.Stat(testBinary); os.IsNotExist(err) {
-		t.Skip("Test binary /tmp/test_pgroup not found - run: gcc -o /tmp/test_pgroup /tmp/test_pgroup.c")
+		t.Skip(
+			"Test binary /tmp/test_pgroup not found - run: gcc -o /tmp/test_pgroup /tmp/test_pgroup.c",
+		)
 	}
 
 	tool, err := NewExecTool("", false)
@@ -1555,8 +1606,16 @@ func TestDetectPtyKeyMode(t *testing.T) {
 		{"rmkx only", "\x1b[?1l\x1b>", PtyKeyModeCSI},
 		{"both smkx first", "\x1b[?1h\x1b=...\x1b[?1l\x1b>", PtyKeyModeCSI},
 		{"both rmkx first", "\x1b[?1l\x1b>...\x1b[?1h\x1b=", PtyKeyModeSS3},
-		{"multiple toggles smkx last", "\x1b[?1h\x1b=...\x1b[?1l\x1b>...\x1b[?1h\x1b=", PtyKeyModeSS3},
-		{"multiple toggles rmkx last", "\x1b[?1l\x1b>...\x1b[?1h\x1b=...\x1b[?1l\x1b>", PtyKeyModeCSI},
+		{
+			"multiple toggles smkx last",
+			"\x1b[?1h\x1b=...\x1b[?1l\x1b>...\x1b[?1h\x1b=",
+			PtyKeyModeSS3,
+		},
+		{
+			"multiple toggles rmkx last",
+			"\x1b[?1l\x1b>...\x1b[?1h\x1b=...\x1b[?1l\x1b>",
+			PtyKeyModeCSI,
+		},
 		{"partial smkx", "\x1b[?1h", PtyKeyModeSS3},
 		{"partial rmkx", "\x1b[?1l", PtyKeyModeCSI},
 	}
diff --git a/pkg/tools/skills_install.go b/pkg/tools/skills_install.go
index 71bfe730b..ffb4b0c52 100644
--- a/pkg/tools/skills_install.go
+++ b/pkg/tools/skills_install.go
@@ -96,7 +96,11 @@ func (t *InstallSkillTool) Execute(ctx context.Context, args map[string]any) *To
 	if !force {
 		if _, err := os.Stat(targetDir); err == nil {
 			return ErrorResult(
-				fmt.Sprintf("skill %q already installed at %s. Use force=true to reinstall.", slug, targetDir),
+				fmt.Sprintf(
+					"skill %q already installed at %s. Use force=true to reinstall.",
+					slug,
+					targetDir,
+				),
 			)
 		}
 	} else {
@@ -142,7 +146,9 @@ func (t *InstallSkillTool) Execute(ctx context.Context, args map[string]any) *To
 					"error":      rmErr.Error(),
 				})
 		}
-		return ErrorResult(fmt.Sprintf("skill %q is flagged as malicious and cannot be installed", slug))
+		return ErrorResult(
+			fmt.Sprintf("skill %q is flagged as malicious and cannot be installed", slug),
+		)
 	}
 
 	// Write origin metadata.
@@ -162,7 +168,10 @@ func (t *InstallSkillTool) Execute(ctx context.Context, args map[string]any) *To
 	// Build result with moderation warning if suspicious.
 	var output string
 	if result.IsSuspicious {
-		output = fmt.Sprintf("⚠️ Warning: skill %q is flagged as suspicious (may contain risky patterns).\n\n", slug)
+		output = fmt.Sprintf(
+			"⚠️ Warning: skill %q is flagged as suspicious (may contain risky patterns).\n\n",
+			slug,
+		)
 	}
 	output += fmt.Sprintf("Successfully installed skill %q v%s from %s registry.\nLocation: %s\n",
 		slug, result.Version, registry.Name(), targetDir)
diff --git a/pkg/tools/skills_search.go b/pkg/tools/skills_search.go
index 2b6cffd38..8f7401dfa 100644
--- a/pkg/tools/skills_search.go
+++ b/pkg/tools/skills_search.go
@@ -17,7 +17,10 @@ type FindSkillsTool struct {
 // NewFindSkillsTool creates a new FindSkillsTool.
 // registryMgr is the shared registry manager (built from config in createToolRegistry).
 // cache is the search cache for deduplicating similar queries.
-func NewFindSkillsTool(registryMgr *skills.RegistryManager, cache *skills.SearchCache) *FindSkillsTool {
+func NewFindSkillsTool(
+	registryMgr *skills.RegistryManager,
+	cache *skills.SearchCache,
+) *FindSkillsTool {
 	return &FindSkillsTool{
 		registryMgr: registryMgr,
 		cache:       cache,
diff --git a/pkg/tools/spawn_status.go b/pkg/tools/spawn_status.go
index 416fd2226..22202d899 100644
--- a/pkg/tools/spawn_status.go
+++ b/pkg/tools/spawn_status.go
@@ -77,10 +77,12 @@ func (t *SpawnStatusTool) Execute(ctx context.Context, args map[string]any) *Too
 		}
 
 		// Restrict lookup to tasks that belong to this conversation.
-		if callerChannel != "" && taskCopy.OriginChannel != "" && taskCopy.OriginChannel != callerChannel {
+		if callerChannel != "" && taskCopy.OriginChannel != "" &&
+			taskCopy.OriginChannel != callerChannel {
 			return ErrorResult(fmt.Sprintf("No subagent found with task ID: %s", taskID))
 		}
-		if callerChatID != "" && taskCopy.OriginChatID != "" && taskCopy.OriginChatID != callerChatID {
+		if callerChatID != "" && taskCopy.OriginChatID != "" &&
+			taskCopy.OriginChatID != callerChatID {
 			return ErrorResult(fmt.Sprintf("No subagent found with task ID: %s", taskID))
 		}
 
diff --git a/pkg/tools/spawn_status_test.go b/pkg/tools/spawn_status_test.go
index 9c772d61a..22b885fb7 100644
--- a/pkg/tools/spawn_status_test.go
+++ b/pkg/tools/spawn_status_test.go
@@ -195,7 +195,12 @@ func TestSpawnStatusTool_TaskID_NonString(t *testing.T) {
 	for _, badVal := range []any{42, 3.14, true, map[string]any{"x": 1}, []string{"a"}} {
 		result := tool.Execute(context.Background(), map[string]any{"task_id": badVal})
 		if !result.IsError {
-			t.Errorf("Expected error for task_id=%T(%v), got success: %s", badVal, badVal, result.ForLLM)
+			t.Errorf(
+				"Expected error for task_id=%T(%v), got success: %s",
+				badVal,
+				badVal,
+				result.ForLLM,
+			)
 		}
 		if !strings.Contains(result.ForLLM, "task_id must be a string") {
 			t.Errorf("Expected type-error message, got: %s", result.ForLLM)
@@ -319,7 +324,10 @@ func TestSpawnStatusTool_SortByCreatedTimestamp(t *testing.T) {
 		t.Fatalf("Both task IDs should appear in output:\n%s", result.ForLLM)
 	}
 	if pos2 > pos10 {
-		t.Errorf("Expected subagent-2 (created first) to appear before subagent-10, but got:\n%s", result.ForLLM)
+		t.Errorf(
+			"Expected subagent-2 (created first) to appear before subagent-10, but got:\n%s",
+			result.ForLLM,
+		)
 	}
 }
 
diff --git a/pkg/tools/spi.go b/pkg/tools/spi.go
index 0ca17e84f..cdf23db86 100644
--- a/pkg/tools/spi.go
+++ b/pkg/tools/spi.go
@@ -69,7 +69,9 @@ func (t *SPITool) Parameters() map[string]any {
 
 func (t *SPITool) Execute(ctx context.Context, args map[string]any) *ToolResult {
 	if runtime.GOOS != "linux" {
-		return ErrorResult("SPI is only supported on Linux. This tool requires /dev/spidev* device files.")
+		return ErrorResult(
+			"SPI is only supported on Linux. This tool requires /dev/spidev* device files.",
+		)
 	}
 
 	action, ok := args["action"].(string)
@@ -124,7 +126,9 @@ func (t *SPITool) list() *ToolResult {
 // parseSPIArgs extracts and validates common SPI parameters
 //
 //nolint:unused // Used by spi_linux.go
-func parseSPIArgs(args map[string]any) (device string, speed uint32, mode uint8, bits uint8, errMsg string) {
+func parseSPIArgs(
+	args map[string]any,
+) (device string, speed uint32, mode uint8, bits uint8, errMsg string) {
 	dev, ok := args["device"].(string)
 	if !ok || dev == "" {
 		return "", 0, 0, 0, "device is required (e.g. \"2.0\" for /dev/spidev2.0)"
diff --git a/pkg/tools/spi_linux.go b/pkg/tools/spi_linux.go
index 9def73662..d03c4ef92 100644
--- a/pkg/tools/spi_linux.go
+++ b/pkg/tools/spi_linux.go
@@ -38,25 +38,46 @@ type spiTransfer struct {
 func configureSPI(devPath string, mode uint8, bits uint8, speed uint32) (int, *ToolResult) {
 	fd, err := syscall.Open(devPath, syscall.O_RDWR, 0)
 	if err != nil {
-		return -1, ErrorResult(fmt.Sprintf("failed to open %s: %v (check permissions and spidev module)", devPath, err))
+		return -1, ErrorResult(
+			fmt.Sprintf(
+				"failed to open %s: %v (check permissions and spidev module)",
+				devPath,
+				err,
+			),
+		)
 	}
 
 	// Set SPI mode
-	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), spiIocWrMode, uintptr(unsafe.Pointer(&mode)))
+	_, _, errno := syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		spiIocWrMode,
+		uintptr(unsafe.Pointer(&mode)),
+	)
 	if errno != 0 {
 		syscall.Close(fd)
 		return -1, ErrorResult(fmt.Sprintf("failed to set SPI mode %d: %v", mode, errno))
 	}
 
 	// Set bits per word
-	_, _, errno = syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), spiIocWrBitsPerWord, uintptr(unsafe.Pointer(&bits)))
+	_, _, errno = syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		spiIocWrBitsPerWord,
+		uintptr(unsafe.Pointer(&bits)),
+	)
 	if errno != 0 {
 		syscall.Close(fd)
 		return -1, ErrorResult(fmt.Sprintf("failed to set bits per word %d: %v", bits, errno))
 	}
 
 	// Set max speed
-	_, _, errno = syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), spiIocWrMaxSpeedHz, uintptr(unsafe.Pointer(&speed)))
+	_, _, errno = syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		spiIocWrMaxSpeedHz,
+		uintptr(unsafe.Pointer(&speed)),
+	)
 	if errno != 0 {
 		syscall.Close(fd)
 		return -1, ErrorResult(fmt.Sprintf("failed to set SPI speed %d Hz: %v", speed, errno))
@@ -117,7 +138,12 @@ func (t *SPITool) transfer(args map[string]any) *ToolResult {
 		bitsPerWord: bits,
 	}
 
-	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), spiIocMessage1, uintptr(unsafe.Pointer(&xfer)))
+	_, _, errno := syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		spiIocMessage1,
+		uintptr(unsafe.Pointer(&xfer)),
+	)
 	runtime.KeepAlive(txBuf)
 	runtime.KeepAlive(rxBuf)
 	if errno != 0 {
@@ -174,7 +200,12 @@ func (t *SPITool) readDevice(args map[string]any) *ToolResult {
 		bitsPerWord: bits,
 	}
 
-	_, _, errno := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), spiIocMessage1, uintptr(unsafe.Pointer(&xfer)))
+	_, _, errno := syscall.Syscall(
+		syscall.SYS_IOCTL,
+		uintptr(fd),
+		spiIocMessage1,
+		uintptr(unsafe.Pointer(&xfer)),
+	)
 	runtime.KeepAlive(txBuf)
 	runtime.KeepAlive(rxBuf)
 	if errno != 0 {
diff --git a/pkg/tools/subagent_tool_test.go b/pkg/tools/subagent_tool_test.go
index 89ac7d4b5..601d3f937 100644
--- a/pkg/tools/subagent_tool_test.go
+++ b/pkg/tools/subagent_tool_test.go
@@ -316,7 +316,11 @@ func TestSubagentTool_ForUserTruncation(t *testing.T) {
 	// ForUser should be truncated to 500 chars + "..."
 	maxUserLen := 500
 	if len(result.ForUser) > maxUserLen+3 { // +3 for "..."
-		t.Errorf("ForUser should be truncated to ~%d chars, got: %d", maxUserLen, len(result.ForUser))
+		t.Errorf(
+			"ForUser should be truncated to ~%d chars, got: %d",
+			maxUserLen,
+			len(result.ForUser),
+		)
 	}
 
 	// ForLLM should have full content
diff --git a/pkg/tools/toolloop.go b/pkg/tools/toolloop.go
index 387813e94..df72301a2 100644
--- a/pkg/tools/toolloop.go
+++ b/pkg/tools/toolloop.go
@@ -64,7 +64,13 @@ func RunToolLoop(
 			llmOpts = map[string]any{}
 		}
 		// 3. Call LLM
-		response, err := config.Provider.Chat(ctx, messages, providerToolDefs, config.Model, llmOpts)
+		response, err := config.Provider.Chat(
+			ctx,
+			messages,
+			providerToolDefs,
+			config.Model,
+			llmOpts,
+		)
 		if err != nil {
 			logger.ErrorCF("toolloop", "LLM call failed",
 				map[string]any{
@@ -148,7 +154,14 @@ func RunToolLoop(
 
 				var toolResult *ToolResult
 				if config.Tools != nil {
-					toolResult = config.Tools.ExecuteWithContext(ctx, tc.Name, tc.Arguments, channel, chatID, nil)
+					toolResult = config.Tools.ExecuteWithContext(
+						ctx,
+						tc.Name,
+						tc.Arguments,
+						channel,
+						chatID,
+						nil,
+					)
 				} else {
 					toolResult = ErrorResult("No tools available")
 				}
diff --git a/pkg/tools/validate_test.go b/pkg/tools/validate_test.go
index e7f4f619a..accff9e3c 100644
--- a/pkg/tools/validate_test.go
+++ b/pkg/tools/validate_test.go
@@ -151,7 +151,10 @@ func TestValidateToolArgs(t *testing.T) {
 			schema: map[string]any{
 				"type": "object",
 				"properties": map[string]any{
-					"color": map[string]any{"type": "string", "enum": []any{"red", "green", "blue"}},
+					"color": map[string]any{
+						"type": "string",
+						"enum": []any{"red", "green", "blue"},
+					},
 				},
 			},
 			args: map[string]any{"color": "red"},
@@ -161,7 +164,10 @@ func TestValidateToolArgs(t *testing.T) {
 			schema: map[string]any{
 				"type": "object",
 				"properties": map[string]any{
-					"color": map[string]any{"type": "string", "enum": []any{"red", "green", "blue"}},
+					"color": map[string]any{
+						"type": "string",
+						"enum": []any{"red", "green", "blue"},
+					},
 				},
 			},
 			args:    map[string]any{"color": "yellow"},
@@ -172,7 +178,10 @@ func TestValidateToolArgs(t *testing.T) {
 			schema: map[string]any{
 				"type": "object",
 				"properties": map[string]any{
-					"color": map[string]any{"type": "string", "enum": []string{"red", "green", "blue"}},
+					"color": map[string]any{
+						"type": "string",
+						"enum": []string{"red", "green", "blue"},
+					},
 				},
 			},
 			args: map[string]any{"color": "green"},
@@ -182,7 +191,10 @@ func TestValidateToolArgs(t *testing.T) {
 			schema: map[string]any{
 				"type": "object",
 				"properties": map[string]any{
-					"color": map[string]any{"type": "string", "enum": []string{"red", "green", "blue"}},
+					"color": map[string]any{
+						"type": "string",
+						"enum": []string{"red", "green", "blue"},
+					},
 				},
 			},
 			args:    map[string]any{"color": "yellow"},
@@ -342,7 +354,11 @@ func TestValidateToolArgs_RegistryIntegration(t *testing.T) {
 	}
 
 	// Extra property — should fail with validation error
-	result = r.Execute(context.Background(), "read_file", map[string]any{"path": "/x", "__inject": true})
+	result = r.Execute(
+		context.Background(),
+		"read_file",
+		map[string]any{"path": "/x", "__inject": true},
+	)
 	if !result.IsError {
 		t.Error("expected validation error for extra property")
 	}
diff --git a/pkg/tools/web_test.go b/pkg/tools/web_test.go
index de6187cfa..2c0de25f7 100644
--- a/pkg/tools/web_test.go
+++ b/pkg/tools/web_test.go
@@ -54,7 +54,8 @@ func TestWebTool_WebFetch_Success(t *testing.T) {
 	}
 
 	// ForUser should contain summary
-	if !strings.Contains(result.ForUser, "bytes") && !strings.Contains(result.ForUser, "extractor") {
+	if !strings.Contains(result.ForUser, "bytes") &&
+		!strings.Contains(result.ForUser, "extractor") {
 		t.Errorf("Expected ForUser to contain summary, got: %s", result.ForUser)
 	}
 }
@@ -75,7 +76,11 @@ func TestWebTool_WebFetch_JSON(t *testing.T) {
 
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -100,7 +105,11 @@ func TestWebTool_WebFetch_JSON(t *testing.T) {
 func TestWebTool_WebFetch_InvalidURL(t *testing.T) {
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -125,7 +134,11 @@ func TestWebTool_WebFetch_InvalidURL(t *testing.T) {
 func TestWebTool_WebFetch_UnsupportedScheme(t *testing.T) {
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -141,7 +154,8 @@ func TestWebTool_WebFetch_UnsupportedScheme(t *testing.T) {
 	}
 
 	// Should mention only http/https allowed
-	if !strings.Contains(result.ForLLM, "http/https") && !strings.Contains(result.ForUser, "http/https") {
+	if !strings.Contains(result.ForLLM, "http/https") &&
+		!strings.Contains(result.ForUser, "http/https") {
 		t.Errorf("Expected scheme error message, got ForLLM: %s", result.ForLLM)
 	}
 }
@@ -150,7 +164,11 @@ func TestWebTool_WebFetch_UnsupportedScheme(t *testing.T) {
 func TestWebTool_WebFetch_MissingURL(t *testing.T) {
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -164,7 +182,8 @@ func TestWebTool_WebFetch_MissingURL(t *testing.T) {
 	}
 
 	// Should mention URL is required
-	if !strings.Contains(result.ForLLM, "url is required") && !strings.Contains(result.ForUser, "url is required") {
+	if !strings.Contains(result.ForLLM, "url is required") &&
+		!strings.Contains(result.ForUser, "url is required") {
 		t.Errorf("Expected 'url is required' message, got ForLLM: %s", result.ForLLM)
 	}
 }
@@ -184,7 +203,11 @@ func TestWebTool_WebFetch_Truncation(t *testing.T) {
 
 	tool, err := NewWebFetchTool(1000, format, testFetchLimit) // Limit to 1000 chars
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -216,7 +239,10 @@ func TestWebTool_WebFetch_Truncation(t *testing.T) {
 	// Text should end with the truncation notice
 	if text, ok := resultMap["text"].(string); ok {
 		if !strings.HasSuffix(text, "[Content truncated due to size limit]") {
-			t.Errorf("Expected text to end with truncation notice, got: %q", text[max(0, len(text)-60):])
+			t.Errorf(
+				"Expected text to end with truncation notice, got: %q",
+				text[max(0, len(text)-60):],
+			)
 		}
 	}
 }
@@ -263,11 +289,13 @@ func TestWebTool_WebFetch_TruncationNotice(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-				w.Header().Set("Content-Type", tt.contentType)
-				w.WriteHeader(http.StatusOK)
-				w.Write([]byte(tt.body))
-			}))
+			server := httptest.NewServer(
+				http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+					w.Header().Set("Content-Type", tt.contentType)
+					w.WriteHeader(http.StatusOK)
+					w.Write([]byte(tt.body))
+				}),
+			)
 			defer server.Close()
 
 			tool, err := NewWebFetchTool(maxChars, tt.format, testFetchLimit)
@@ -291,7 +319,11 @@ func TestWebTool_WebFetch_TruncationNotice(t *testing.T) {
 			}
 
 			if !strings.HasSuffix(text, truncationNotice) {
-				t.Errorf("expected text to end with %q, got suffix: %q", truncationNotice, text[max(0, len(text)-60):])
+				t.Errorf(
+					"expected text to end with %q, got suffix: %q",
+					truncationNotice,
+					text[max(0, len(text)-60):],
+				)
 			}
 
 			if truncated, ok := resultMap["truncated"].(bool); !ok || !truncated {
@@ -360,7 +392,11 @@ func TestWebFetchTool_PayloadTooLarge(t *testing.T) {
 	// Initialize the tool
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	// Prepare the arguments pointing to the URL of our local mock server
@@ -380,7 +416,8 @@ func TestWebFetchTool_PayloadTooLarge(t *testing.T) {
 	// Search for the exact error string we set earlier in the Execute method
 	expectedErrorMsg := fmt.Sprintf("size exceeded %d bytes limit", testFetchLimit)
 
-	if !strings.Contains(result.ForLLM, expectedErrorMsg) && !strings.Contains(result.ForUser, expectedErrorMsg) {
+	if !strings.Contains(result.ForLLM, expectedErrorMsg) &&
+		!strings.Contains(result.ForUser, expectedErrorMsg) {
 		t.Errorf("test failed: expected error %q, but got: %+v", expectedErrorMsg, result)
 	}
 }
@@ -533,7 +570,11 @@ func TestWebTool_WebFetch_HTMLExtraction(t *testing.T) {
 
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -718,7 +759,13 @@ func TestWebTool_WebFetch_PrivateHostAllowedByCIDRWhitelist(t *testing.T) {
 	defer server.Close()
 
 	host, _ := serverHostAndPort(t, server.URL)
-	tool, err := NewWebFetchToolWithConfig(50000, "", format, testFetchLimit, []string{singleHostCIDR(t, host)})
+	tool, err := NewWebFetchToolWithConfig(
+		50000,
+		"",
+		format,
+		testFetchLimit,
+		[]string{singleHostCIDR(t, host)},
+	)
 	if err != nil {
 		t.Fatalf("Failed to create web fetch tool: %v", err)
 	}
@@ -753,7 +800,10 @@ func TestWebTool_WebFetch_PrivateHostAllowedForTests(t *testing.T) {
 	})
 
 	if result.IsError {
-		t.Errorf("expected success when private host access is allowed in tests, got %q", result.ForLLM)
+		t.Errorf(
+			"expected success when private host access is allowed in tests, got %q",
+			result.ForLLM,
+		)
 	}
 }
 
@@ -973,7 +1023,11 @@ func TestIsPrivateOrRestrictedIP_Table(t *testing.T) {
 func TestWebTool_WebFetch_MissingDomain(t *testing.T) {
 	tool, err := NewWebFetchTool(50000, format, testFetchLimit)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	ctx := context.Background()
@@ -995,9 +1049,19 @@ func TestWebTool_WebFetch_MissingDomain(t *testing.T) {
 }
 
 func TestNewWebFetchToolWithProxy(t *testing.T) {
-	tool, err := NewWebFetchToolWithProxy(1024, "http://127.0.0.1:7890", format, testFetchLimit, nil)
+	tool, err := NewWebFetchToolWithProxy(
+		1024,
+		"http://127.0.0.1:7890",
+		format,
+		testFetchLimit,
+		nil,
+	)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	} else if tool.maxChars != 1024 {
 		t.Fatalf("maxChars = %d, want %d", tool.maxChars, 1024)
 	}
@@ -1008,7 +1072,11 @@ func TestNewWebFetchToolWithProxy(t *testing.T) {
 
 	tool, err = NewWebFetchToolWithProxy(0, "http://127.0.0.1:7890", format, testFetchLimit, nil)
 	if err != nil {
-		logger.ErrorCF("agent", "Failed to create web fetch tool", map[string]any{"error": err.Error()})
+		logger.ErrorCF(
+			"agent",
+			"Failed to create web fetch tool",
+			map[string]any{"error": err.Error()},
+		)
 	}
 
 	if tool.maxChars != 50000 {
@@ -1017,7 +1085,13 @@ func TestNewWebFetchToolWithProxy(t *testing.T) {
 }
 
 func TestNewWebFetchToolWithConfig_InvalidPrivateHostWhitelist(t *testing.T) {
-	_, err := NewWebFetchToolWithConfig(1024, "", format, testFetchLimit, []string{"not-an-ip-or-cidr"})
+	_, err := NewWebFetchToolWithConfig(
+		1024,
+		"",
+		format,
+		testFetchLimit,
+		[]string{"not-an-ip-or-cidr"},
+	)
 	if err == nil {
 		t.Fatal("expected invalid whitelist entry to fail")
 	}
@@ -1173,7 +1247,11 @@ func TestWebTool_TavilySearch_RangeMapping(t *testing.T) {
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(map[string]any{
 			"results": []map[string]any{
-				{"title": "Recent result", "url": "https://example.com/recent", "content": "snippet"},
+				{
+					"title":   "Recent result",
+					"url":     "https://example.com/recent",
+					"content": "snippet",
+				},
 			},
 		})
 	}))
@@ -1303,7 +1381,10 @@ func TestWebFetchTool_CloudflareChallenge_RetryFailsToo(t *testing.T) {
 
 	// Should not be an error — the retry response is used as-is (403 is a valid HTTP response)
 	if result.IsError {
-		t.Fatalf("expected non-error result even when retry is also blocked, got: %s", result.ForLLM)
+		t.Fatalf(
+			"expected non-error result even when retry is also blocked, got: %s",
+			result.ForLLM,
+		)
 	}
 	// Status in the JSON result should reflect the 403
 	if !strings.Contains(result.ForLLM, "403") {
@@ -1468,7 +1549,10 @@ func TestWebTool_GLMSearch_Success(t *testing.T) {
 			t.Errorf("Expected Content-Type application/json, got %s", r.Header.Get("Content-Type"))
 		}
 		if r.Header.Get("Authorization") != "Bearer test-glm-key" {
-			t.Errorf("Expected Authorization Bearer test-glm-key, got %s", r.Header.Get("Authorization"))
+			t.Errorf(
+				"Expected Authorization Bearer test-glm-key, got %s",
+				r.Header.Get("Authorization"),
+			)
 		}
 
 		var payload map[string]any
@@ -1534,14 +1618,21 @@ func TestWebTool_GLMSearch_RangeMapping(t *testing.T) {
 			t.Fatalf("failed to decode payload: %v", err)
 		}
 		if payload["search_recency_filter"] != "oneMonth" {
-			t.Fatalf("expected search_recency_filter=oneMonth, got %v", payload["search_recency_filter"])
+			t.Fatalf(
+				"expected search_recency_filter=oneMonth, got %v",
+				payload["search_recency_filter"],
+			)
 		}
 
 		w.Header().Set("Content-Type", "application/json")
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(map[string]any{
 			"search_result": []map[string]any{
-				{"title": "Recent GLM Result", "content": "snippet", "link": "https://example.com/glm-range"},
+				{
+					"title":   "Recent GLM Result",
+					"content": "snippet",
+					"link":    "https://example.com/glm-range",
+				},
 			},
 		})
 	}))
@@ -1573,14 +1664,21 @@ func TestWebTool_BaiduSearch_RangeMapping(t *testing.T) {
 			t.Fatalf("failed to decode payload: %v", err)
 		}
 		if payload["search_recency_filter"] != "week" {
-			t.Fatalf("expected search_recency_filter=week for day fallback, got %v", payload["search_recency_filter"])
+			t.Fatalf(
+				"expected search_recency_filter=week for day fallback, got %v",
+				payload["search_recency_filter"],
+			)
 		}
 
 		w.Header().Set("Content-Type", "application/json")
 		w.WriteHeader(http.StatusOK)
 		json.NewEncoder(w).Encode(map[string]any{
 			"references": []map[string]any{
-				{"title": "Recent Baidu Result", "url": "https://example.com/baidu", "content": "snippet"},
+				{
+					"title":   "Recent Baidu Result",
+					"url":     "https://example.com/baidu",
+					"content": "snippet",
+				},
 			},
 		})
 	}))